Trade Policy and Economic Welfare

TRADE POLICY AND ECONOMIC WELFARE

This page intentionally left blank

Trade Policy and Economic Welfare Second Edition

W, MAX CORDEN

CLARENDON PRESS • OXFORD 1997

This book has been printed digitally and produced in a standard specification in order to ensure its continuing availability

OXFORD UNIVERSITY PRESS

Great Clarendon Street, Oxford OX2 6DP Oxford University Press is a department of the University of Oxford, It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan South Korea Poland Portugal Singapore Switzerland Thailand Turkey Ukraine Vietnam

Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York © Oxford University Press 1997 The moral rights of the author have been asserted Database right Oxford University Press (maker) Reprinted 2007 All rights reserved, No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover And you must impose this same condition on any acquirer ISBN 978-0-19-829223-4

In Memory of My Father

This page intentionally left blank

Preface to the Second Edition This book deals with the normative theory of trade policy, primarily from the point of view of a single country. The first edition of this book has had a long and gratifying life, finally going out of print in 1994, after twenty years. Yet its topics and methodology are still very relevant, even though updating has been required, and even though there has been much trade liberalization since 1974, especially by developing countries, For the second edition I have added three new chapters, on trade policy and the environment (Chapter 13), on strategic trade and industrial policy (Chapter 14), and on the relationship between trade policy, the exchange rate, and the current account (Chapter 15). Much of Chapter 13 is a direct application to a popular subject of the basic methodology of this book. Chapter 14 deals with a topic that has received much attention in the professional literature in the 1980s, and even more—to an excessive extent—in nonprofessional policy discussion, Chapter 15 draws heavily on many writings of mine since 1974 on trade policy and the macroeconomy, especially the exchange rate. Coming to the other chapters, taken from the original book, the key chapter is Chapter 2, which provides the framework for the whole book, Looking at these chapters now, after twenty years, and in the light of current discussion, the other chapter that seems to me most important and that I should like every teacher and student to study carefully is Chapter 8 on the infant industry argument. I have left this chapter as it was (apart from a few footnote changes) since it is totally relevant now, and also quite comprehensive. Many current arguments for protection, including some discussed in Chapter 14 on strategic trade and industrial policy, can be recognized in that chapter Chapter 5, on income distribution, contains the idea of the "conservative social welfare function' which has received attention in the literature on the political economy of protection and is one reason (possibly the main one in later years) why the book has been much cited in the economic journals. Chapters 7,9 and 10 deal with topics that received little attention at the time the original book was written but that have become prominent since, For this edition I have compressed the original book, eliminating completely two chapters (on employment and industrialization, and on cost benefit analysis), and also cutting out various passages and sections which dealt with fine points or seem not very interesting now. 1 have made some improvements in exposition, but the basic theory is unchanged. Also, I have updated numerous footnotes. In several passages of the original book the language was sexist. The editor and I have struggled to make the language more inclusive, but only succeeded

viii

Preface to the Second Edition

In making difficult passages more obscure. Hence, with apologies, monopolists have remained male while the private risk-taker in section 10.3 has become female, This edition was primarily written and edited at the School of Advanced International Studies (SAIS) of The Johns Hopkins University, but completed during a visit to the Australian National University (where many of the ideas of the first edition were conceived). I am grateful to both universities for providing me with excellent facilities. I owe a particular debt to Mrs Flora Paoli, my secretary at SAIS, who worked hard, and most skillfully with me on editing the 1974 book to fit in with this edition, and on preparing the new book for the publisher, I am also indebted for help and criticisms to Kym Anderson {Chapter 13), and to George Fane and Ben Smith (Chapter 15).

WMC Washington DC January 1996

Preface to the First Edition This book is a companion to The Theory of Protection, published in 1971. While a reader need not be familiar with the earlier book to follow the present one, the two books together result from my work over many years on the theory and practice of trade policies and the principles of government intervention, The present book has turned out larger than originally intended, and perhaps says more on the subject than every reasonably interested reader would want to read. Thus I should like to stress that Chapters 2 and 3 present much of the basic method of analysis, while the other chapters can be regarded as being more or less self-contained: the reader can select from them according to his interests. My debts are many. My late father, to whom I have dedicated this book, first kindled my interests in the general field of trade restrictions and economic policy. I began work on tariffs in 1955, publishing an article on 'The Calculation of the Cost of Protection' in the Economic Record in 1957. From 1958 to 1967 I was at Melbourne University and the Australian National University, and developed many of my ideas while working on Australian tariff policy. Thus I owe considerable debts to my colleagues and friends of those days, and to the two universities which allowed me time and facilities. I continued working on the subject when I came to Nuffield College in 1967, and benefited from an environment in which trade policies of developing countries were much under discussion, It will be obvious to any reader where my principal intellectual debts are owed: to James Meade's Trade and Welfare (1955), and to Harry Johnson's article 'Optimal Trade Intervention in the Presence of Domestic Distortions* (1965). Much of the book is a variation upon the themes of these two works, though I hope I have carried their analyses a little further. My debt to Harry Johnson is much greater than through the paper just cited. During the period that I have worked in this field he has completely dominated it, as is evident from numerous citations in this book. I have built on his work, sometimes criticized or gone beyond it, but always benefited from it. Not least, he has helped me with constant encouragement. In addition, I have been influenced by, and have built on, the work of many other contributors to the literature of trade theory and policy during recent years, particularly Jagdish Bhagwati, and also Bela Balassa, Murray Kemp, Ronald Jones, and the late V. K. Ramaswami, I owe a great deal to many good friends who, at some stage, read various parts of the manuscript and led me to make many changes, to clarify points, eliminate errors, and rearrange chapters. The whole manuscript was read by John Black, Rattan Bhatia, John Martin, and Jean van der Mensbrugghe, John

x

Preface to the First Edition

Flemming particularly helped me with some problems in Chapters 4 and 10 (especially the discussion of risk avoidance), and 1 am also indebted for help or criticisms to Heinz Arndt (Chapter 11), Stanley Engerman (Chapter 5), Ronald Jones (Chapter 5), Vijay Joshi (Chapter 9), P. R, Narvekar, Maurice Scott (Chapter 9), Richard Snape (Chapter 7), K. P. Teh and David Wall. Graduate students of Oxford University, Princeton University and the University of Minnesota also innocently influenced this book with their often very perceptive, if disconcerting, questions. The book was mainly written in Nuffield College, but also in the Woodrow Wilson School of Princeton University; in addition, I wrote little bits of it in the University of Minnesota, Monash University and La Trobe University, and to all these institutions I am much indebted. The manuscript was typed competently and cheerfully by Penny Sylvester at Nuffield College and by Mary Leksa at the Woodrow Wilson School. I am deeply grateful to them for their patience and efficiency. Finally, my wife and daughter again put up with a husband and father still dreaming about theories, still drawing diagrams, still hammering away at a typewriter. Nuffield College, Oxford January 1974

Acknowledgements Section 7.5 is a revised version of 'The Efficiency Effects of Trade and Protection,' which appeared in I. A. McDougall and R, H, Snape (eds,), Studies in International Economics: Manash Conference Papers, NorthHolland, Amsterdam 1970. An earlier version of Chapter 9 (excluding section 9.3) appeared in Luis di Marco (eel), International Economics and Development, Academic Press, New York 1972. Much of Chapter 11 is based on 'Protection and Foreign Investment,' The Economic Record, 43, June 1967. This chapter also includes some passages from "The Multinational Corporation and International Trade Theory,' in John Dunning (ed.), The Multinational Corporation and Economic Analysis, Allen and Unwin, London 1974. Section 14.1, is a revised version of 'Strategic Trade Policy,' in David Greenaway et al. (eds.), Companion to Contemporary Economic Thought, Routledge, London 1991. 1 am indebted to the editors and publishers of all these publications for permission to make use of this work.

This page intentionally left blank

Contents 1. INTRODUCTION

1.1 Three Stages of Thought 1.2 The Scope of the Book 1.3 Readers' Guide 2. THE THEORY OF DOMESTIC DIVERGENCES

2.1 A Simple Model: Marginal Divergence and the Optimum Subsidy 2.2 By-product Distortion Effect of Tariff 2.3 Elaborations of the Simple Analysis 2.4 The Second-best Optimum Tariff 2.5 Marginal Divergence in Factor Cost 2.6 General Equilibrium and Many Commodities 2.7 Home-market Bias 2.8 The Hierarchy of Policies 2.9 Trade Divergences Appendix 2.1: The Theory of Domestic Divergences in Two-product Geometry Appendix 2.2: The Theory of Domestic Divergences: Notes on the Literature 3. THE FOUR ASSUMPTIONS OF THE THEORY OF DOMESTIC DIVERGENCES

3.1 By-product Distortion Costs of Revenue-rats ing: Subsidies Remain First-best 3.2 Collection Costs 3.3 By-product Costs of Revenue-raising: The Need for 'Trading-off 3.4 Subsidy Disbursement Costs 3.5 Income Distribution Effects and the Theory of Domestic Divergences 3.6 Income Redistribution Not Costless 3.7 Illusions and an Imperfect World

1

2 4 6

7 7 9 11 14 15 17 18 21 23 26 32 33 34 35 36 37 39 40 42

xiv

Contents

4. TRADE TAXES AS SOURCES OF GOVERNMENT R E V E N U E

4.1 4.2 4.3 4.4

Some Simple Principles of Optimum Taxation The Importance of Collection Costs Smuggling and Underinvoicing The Optimum Tariff Structure: The Marginal Cost of Raising Revenue 4.5 Export Taxes 4.6 The Declining Importance of Trade- Tax Revenue in the Process of Economic Development 5. PROTECTION AND INCOME DISTRIBUTION

5.1 5.2 5.3 5.4 5.5

The Effects of Protection on Income Distribution The Stolper-Samuelson Model The Conservative Social Welfare Function Senescent Industry Protection and Adjustment Assistanc interregional Income Distribution

6. THE TERMS OF TRADE ARGUMENT FOR TAXES ON TRADE

6.1 The Optimum Trade Tax in the Two-sector Model 6.2 Multicommodity Model 6.3 Private Monopoly 6.4 Foreign Reaction and Retaliation 6.5 Growth and the Terms of Trade 6.6 Practical Significance of the Tenns-of-tmde Argument Appendix 6.1: The Optimum Export Tax and Optimum Tariff 1. MONOPOLY, MARKET STRUCTURE AND ECONOMIES OF SCALE

7. / 7.2 7.3 7.4 7.5 7.6 7.7

The Effects of Protection on Monopoly Profits Fragmentation of Production Tariffs and Market Structure Made-to-measure Tariffs The X-efficiency Effects of Protection and Monopoly The Rent-seeking Society Dumping

45 46 49 51 53 5?

58 61 61 64 72 76 78 81

82 88 89 90 93 95

100 105 105 111 114 117 120 126 128

Contents 8. THE INFANT INDUSTRY A R G U M E N T

8.1 8.2 8.3 8.4 8.5 8.6 8.7

Dynamic Internal Economies Dynamic External Economies The Mill-Bastable Test Wide Application of the Infant Industry Argument The Problem of Investment Coordination Economies of Scale and the Pseudo-Infant Industry Argument Research and Development, and the Protection of Advanced Technology Industries

9. PROTECTION AND CAPITAL ACCUMULATION

9, / The Real Income Effect of Protection on Savings and Investment 9.2 Savings and the Income Distribution Effects of Protection 9.3 Cheap Imported Capital Goods, Investment and Employment 9.4 Consumption Postponement and the Demonstration Effect 9.5 The Inducement to Invest and Backward Linkage 9.6 Learning and the Rate of Growth 10. SOME DYNAMIC ASPECTS OF TRADE POLICY

10.1 10.2 10.3 10.4 10.5

The Anticipation of Price Changes Fluctuations in Export Earnings Risk Avoidance The Cost of Protection and the Rate of Growth The 'Dynamic" Effects of Trade on Knowledge

11. PROTECTION AND FOREIGN INVESTMENT

XV

139 140 146 151 153 154 156 158 162 162 165 167 174 175 179 181 182 183 185 188 190 193

11.1 The Effect of Protection on Foreign Capital Inflow 11.2 Taxation of Foreign Investment

194 195 11.3 Tariffs as Second-best Taxes'or Subsidies on Foreign Capital 205 11.4 The Effects of a Given Tariff System on the Gains or Losses from Foreign Investment 11.5 Multinational Corporations

207 208

12. OPTIMAL TARIFF DISMANTLING

215

12.1 Some Tariffs are not Alterable 12.2 Concertina Method

216 217

xvi

Contents 12.3 Across-the-Board Reduction 12.4 Tariff Dismantling and the Exchange Rate

13. THE ENVIRONMENT AND TRADE POLICY 13.1 Trade Liberalization and the Environment: Small Country Case 13.2 Environmental Policy Interaction: Multicountry Case but no Tramborder Pollution 13.3 Tramborder Externalities and Global Public Goods 13.4 Capital Mobility and Capital Controls 13.5 Summary: The Role of Trade Policy 14. STRATEGIC TRADE POLICY

14.1 Oligopoly and Strategic Trade Policy 14.2 Industrial Policy 14.3 Japanese Industrial Policy: What Difference Did It Make? 15. THE EXCHANGE RATE, TRADE POLICY, AND THE CURRENT ACCOUNT 15. 1 Trade Liberalization Requires Depreciation 15.2 The Basic Theory 15.3 Exchange Rate Protection 15.4 Capital Inflow and the Real Appreciation Problem 15.5 Real Wage Rigidity and a Balance-of-payments Problem 15.6 How Protection Affects the Current Account 15.7 The Fixed Exchange Rate Regime: What Role for Protection? 15.8 Summary

218 219

222 222 231 236 240 242 244 244 255 258 262 262 263 267 268 271 273 275 277

16. CONCLUSION

278

Bibliography Index of Names Index of Subjects

285 297 299

1

Introduction The aim of this book is to expound, review, and develop the normative theory of trade policy. The concern is mainly with tariffs, but all forms of government policies affecting trade, notably import quotas, export taxes, and taxes and subsidies on the production of particular products or industries, or on the use of factors, come within the scope of the study. More generally, its subject matter is really the principles of government intervention in the economy at the microlevel, especially government subsidization of private industry. While the book rests heavily on international trade theory, it goes beyond some of the theory's common assumptions—two goods, perfect competition, fixed factor supplies, and availability of nondistorting lump-sum transfers. It puts particular emphasis on income distribution effects of trade policies and on the fiscal implications of various kinds of government intervention, It is hardly necessary to stress the significance of the subject matter. The issue of free trade versus protection dominated much economic discussion in Britain, France, Germany, and the United States in the nineteenth century. As this was being written in 1973, the United States appeared to be going through a protectionist phase, and that could still be said in 1995! The subject is a matter of daily newspaper discussion. Tariffs are now less important in developed countries than they used to be, and in Western Europe agricultural industries are now the main beneficiaries of tariff protection. But there are also many forms of nontariff intervention, for example, through preferences to local producers in government procurement, and through subsidies to so-called 'advanced technology' industries.1 It is for developing countries that the issue of protection, whether by tariffs or by import quotas, has been really important. In many developing countries tariffs and import quotas are the principal means by which government policy influences the pattern of economic development, and rates of protection for manufacturing industries were until the mid-1980s—and in some cases still are—very high.2 For this reason much of the book is directed to illuminating ' Trade policies of all major countries are described in International Monetary Fund (1992, 1994) and in similar earlier IMF surveys. On US trade policies, see Kruegcr (1995). An early study of nontariff distortions is Baldwin {1970). Detailed information on many countries* trade policies can be found in the Trade Policy Review reports published by the World Trade Organization and its predecessor, the GATT, 2 The empirical literature on protection in developing countries is vast. See especially Papageorgiou ct at, (1991), and the associated World Bank country studies; and Dean et al. (1994), which describes the current (early 1990s) situation in thirty-two countries. Earlier descriptions— before the great liberalization of the 1980s—are in Little et al, (1970) and Balassa et al. (1971).

2

Introduction

economic policies in developing countries and issues that have been debated by economists advising these countries.

1.1 Three Stages of Thought The transformation of the normative theory of trade policy that took place in the 1950s and 1960s dominates the whole of this book, and its essentials are expounded and somewhat further developed in the next chapter, while various difficulties and refinements are discussed in Chapter 3. One could, rather crudely, classify the evolution of modern postmercantilist economic thought about tariffs and other trade-restricting devices into three stages, In the first stage, the gains from trade and, more specifically,, the benefits from completely free trade, came to be appreciated. The great tool of analysis to emerge was the law of comparative costs, This stage is, above all, associated with the names of Adam Smith and David Ricardo, but it dominated economic thought in Britain for the whole of the nineteenth century, and found clear expression in that 'Samuelson's Economics" of the latter half of the nineteenth century, John Stuart Mill's Principles of Political Economy, The relevant point is that the case for free trade was developed simultaneously with the case for laissez-faire. Indeed, the case for free trade was really a special case of the argument for laissez-faire. Gradually, more and more reasons came to be seen why laissez-faire may not lead to an optimum for a country. Perfect competition does not necessarily rule, there may not be full employment, the income distribution that is yielded by the laissez-faire solution may not be a desirable or just one, necessary structural changes may not take place, and so on. Parallel with the qualifications to laissez-faire, more and more qualifications to the argument for complete free trade were developed. This has been the second stage of thought. One of the arguments for tariffs, the infant industry argument, was even sanctified by the approval of John Stuart Mill. Another, the argument that a country can improve its terms of trade at the expense of other countries and so obtain net gains by some degree of trade restriction, was first systematically expounded by Bickerdike early this century, and was much refined and emphasized later, though indeed the basic idea can also be found in Mill's Principles, The result was that, until the 1960s, the whole subject of free trade and protection was typically expounded in the textbooks as follows. First, the case for free trade, resting on the law of comparative costs, would be developed. Then, in subsequent chapters, various 'arguments for protection*—that is, for possible qualifications to the basic case—would be presented. Some of the more popular arguments would be found fallacious (such as the 'pauper labor' argument) but others would be found valid provided various specified and possibly

Three Stages of Thought

3

not unrealistic assumptions held. The arguments for protection emerged pari passu with the qualifications to the case for laissez-faire, and since there are numerous qualifications to the latter, there are also numerous arguments for protection, As the free trade case was more and more overlaid by numerous sophisticated protectionist arguments, faith in the gains from relatively free trade was gradually lost by many students of economics, and in many countries, especially the developing ones. The exception to this generalization could be found mainly in some countries of Western Europe, where a strong belief in the market economy revived as a result of postwar successes, But during the 1950s the case for free or freer trade was at a discount in most of the countries. Laissez-faire had failed to develop them, so it was argued; hence free trade had failed. In the third stage of thought the link between the case for free trade and the case for laissez-faire was broken. One can believe that there are many reasons for the government to intervene in the economy—to maintain full employment, to bring about a desirable distribution of income, to adjust resource allocation and consumption patterns in the light of external economies and diseconomies, and so on—and yet one can also believe that, broadly, Tree trade is best.' The point is simply that intervention in trade may not be the best way of dealing with the various problems mentioned since they are all essentially caused by 'domestic distortions' rather than 'trade distortions.' The best way may be to deal with them in some direct 'first-best' way, and at the same time to allow trade to flow freely. Of course, if, for some reason, the first-best ways of correcting the distortions are not used, or cannot be used, then trade intervention may still be better than nothing. This approach has been called the theory of domestic distortions, though the term 'distortions' is somewhat loaded and hence will be used in a more limited sense in this book, The new approach radically transformed the theory and has three implications. First, it restores the argument for free or freer trade, though always subject to some exceptions, as we shall see. Second, it focuses attention on the choices between different policies—such as the choice between tariff and subsidy—as distinct from the choice between trade intervention and no intervention. Third, it provides a greatly improved method of analyzing the effects of trade policies, Essentially, the new approach, and the techniques of analysis which it uses, stem from J, E. Meade's Trade and. Welfare. The most important single proposition was highlighted and expounded systematically in an article by Harry G. Johnson, 'Optimal Trade Intervention in the Presence of Domestic Distortions.'3 Appendix 2.2 gives some fuller notes on the history of these ideas 3

Meade(1955a); Johnson (1965).

4

Introduction

and more references. In this book an attempt is made to expound and develop the new approach, especially with respect to its practical applications, to show its wide applicability, and at the same time to point out its limitations.

1.2 The Scope of the Book This book is addressed to professional economists, especially the more practical minded among them, and to students who have some familiarity with international trade theory. Familiarity with elementary welfare economics would also be helpful. It does not begin at the beginning. The beginning, one would think, is a simple exposition of the positive theory of trade policy, and especially of what a tariff and—of more interest—a structure of tariffs, export taxes, and so on, may do to production and consumption patterns and to factor prices. This was expounded in the author's earlier book, The Theory of Protection (197la), though the degree of detail developed there is not a prerequisite for the present book. Some elementary familiarity with the positive theory as it can be derived from any modern basic text on international economics will suffice. Upon positive theory rests normative theory—the theory of policy. The foundation of the normative theory is the familiar argument that countries can mutually gain from trade. For each country the opportunity to trade extends its choice—its frontier of consumption (and investment) possibilities (Samuelson, 1962). This simple idea rests essentially on the law of comparative costs or is a modern development of it. The proposition that there are potentially some gains from trade to a country is perhaps helpful in disposing of some elementary fallacies, but it is not in itself sensational. Some might even say that it is obvious. The more important proposition is that, given certain assumptions, free trade is optimal. The assumptions upon which this proposition rests are notably that there is perfect competition, that all goods and factor services pass through the market, that there are no distorting taxes and other interventions, and that a country cannot affect its terms of trade. In that case free trade will, for any pair of goods, automatically bring about equality between the marginal rate of transformation in domestic production, the marginal rate of substitution in domestic consumption, and the marginal rate of transformation through trade (or, to be more precise, will ensure the absence of certain inequalities). The fulfillment of all the marginal conditions will suffice to ensure "Pareto-optimality,' A concern with the latter implies that domestic income distribution effects are ignored. These elementary or basic propositions, which can be found in many general economic textbooks, and all textbooks of international trade theory, are not expounded here.4 The aim, rather, is to clarify or elaborate what texts sometimes 4 The literature on the existence of competitive equilibrium and the requirements for its •Pareto-optimality' is relevant. See Quirk and Saposnik (1968. Ch. 4). If a competitive equilibrium

Scope of the Book

5

leave obscure or inadequately interpreted, and to carry on where the texts leave off. Further, some issues that have not yet systematically entered the textbooks are pursued. All aspects of trade policy will not be covered. The book does not deal at all with country (as distinct from product) discrimination in tariffs—with preferences, customs unions, and various forms of regional integration. This topic is a large one of its own, and there are good books and surveys available.5 Also, it does not deal with the theory of the measurement of protection—notably the concept of the average tariff, the measurement of the cost of protection and of effective protection,6 Rather, a number of topics are selected—mainly on the basis of their policy relevance -and are analyzed thoroughly. In addition, this is not a book of rigorous general equilibrium theory. Readers who are interested in rigorous formal theory where everything ties into a single general equilibrium system which is fully expounded and all assumptions are always carefully specified should stop here. Every economist must always have a general equilibrium system in his mind and fit a particular partial analysis into it. The partial analysis is unlikely to tell the whole story; it will be a piece of a jigsaw, and the approach here is sometimes to look at one or two pieces on their own with some care. But it should never be forgotten that it is only a piece. Common sense and an appreciation of the limitations of partial analysis must be used in interpreting some of the piecemeal analysis. Finally, and most importantly, this book is only theory. It is meant to be relevant theory, guided in its selection of issues and variables by the author's knowledge of the world of economic affairs, but it is not a direct guide to economic policy in particular countries. At particular stages, in particular chapters or sections, arguments about optimal policies are developed subject to specified assumptions, and perhaps subject to some unspecified ones. But these only say: if the following assumptions hold-—and ignoring many matters discussed in other chapters—the following is 'optimal' policy. The author has views about economic policy in some countries at some times, but these views are influenced not only by an appreciation of the relevant theory but also by a knowledge of the relevant facts and likely reactions to policies, and by value judgments about such matters as income distribution. Theory is vital, but it is not enough. Theory does not 'say*—as is often asserted by the ill-informed or the badly taught—that 'free trade is best.* It says that, exists, which requires (among other things) 'convexity,' and if the closed-economy Parctooptimality requirements are fulfilled, then free trade will (broadly) be Pareto-optima! from a world point of view—and with given terms of trade, from a national point of view. See also Section 5.3 below for further discussion of 'Pareto-optimality.' 5 On customs union theory, see Meade (1955c), Lipsey (1960), Vanek (1965), Krauss (1972), Corden (1984a) and Pomfret (1988). A comprehensive and up-to-date review of all aspects of regional integration is in de Melo and Panagariya (1993). 6 On the cost of protection, see Corden (1957a) and Johnson (1960). On the average tariff, see Corden (1966a). On effective protection, see Corden (1966b, 1971 a), Balassa et at. (1971), and Ethier(I977).

6

Introduction

given certain assumptions, it is "best.' Appreciation of the assumptions under which free trade or alternatively any particular system of protection or subsidization is best, or second-best, third-best, and so on, is perhaps the main thing that should come out of this book.

1.3 Readers'Guide The plan of the book is simple. The most important chapter, setting the theme for the whole book, is the next one. Almost as central is Chapter 3, which qualifies the arguments of Chapter 2, Chapter 16 summarizes the main 'firstbest' results of earlier chapters, listing the circumstances when the use of tariffs might be first-best for a country. It also contains some broader reflections about political economy. Every other chapter can essentially stand on its own and so can be read on its own. "There are, of course, inevitable cross-references. Chapter 5, dealing with the income distribution effects of protection, might be regarded as central to the book, along with Chapters 2 and 3. It introduces the concept of the conservative social welfare function. Chapter 4 (revenue effects) and Chapter 6 (terms of trade) deal in detail with two standard topics. A number of later chapters, notably Chapter 7 on monopoly, Chapter 10 on dynamic aspects, and parts of Chapter 8, opened up some areas that, when they were written in 1973, were essentially new for international trade theory. Chapter 8 (infant industry argument), Chapter 9 (capital accumulation), Chapter 10 (dynamic aspects), and Chapter 11 (foreign investment) are concerned with growth, or at least with models in which the factor supply is not given. Finally, Chapters 13 (environment), 14 (strategic trade policy), and 15 (exchange rate and the current account) are the chapters that were written in 1995.

2

The Theory of Domestic Divergences 2.1 A Simple Model: Marginal Divergence and the Optimum Subsidy A very simple diagram can be used to illustrate the argument that in certain circumstances a subsidy is to be preferred to a tariff as a method of protection,! In Figure 2,1 the quantity of a particular product, an importable product, is shown along the horizontal axis, and its price along the vertical axis. The domestic demand curve for the product is DD', the foreign supply curve for imports of the product is PP' and the supply curve of the domestic importcompeting producers is GG1, In the absence of any tariffs, subsidies, or other intervention, domestic production would be OA, demand would be OB, and the excess of demand over production, AB, would be imports.

FIG, 2,1

The next, and crucial, step is to introduce a marginal divergence between private and social cost. This will be the justification for some kind of government intervention in the economy. GG' can be regarded as indicating the marginal private cost of production for various levels of output. But the social cosi is assumed to be less. We can imagine that external economies of some kind attach to production of this product: there are social benefits that are not taken into account in the private cost calculus. The value of these benefits should be subtracted from the costs from a social point of view. Hence we obtain a curve

i

See Appendix 2,2 for some notes on the history of these ideas and full references.

8

Theory of Domestic Divergences

showing the marginal social cost of production at various outputs, namely HH'. Its general characteristic is that it is below GG', It is of course possible that external diseconomies attach to this particular type of output; in that case HH' would be above GG'. Two more assumptions must be made at this stage, (i) The demand curve, £>£>', indicates not only the private but also the social value at the margin of various quantities consumed. Thus there is no marginal divergence on the demand side, (ii) The price of imports, OF, facing private producers and consumers correctly indicates not only the private but also the social cost of imports. The price is assumed to be unaffected by the quantity of imports the country wishes to purchase, so that average and marginal social costs of imports coincide. This is the small country assumption. The assumption that OP correctly indicates the marginal social cost of imports also means, among other things, that the exchange rate is in equilibrium. We can assume that any reductions in imports of this particular product have only a minor effect on the balance of payments, so that exchange rate alteration would not be required. We shall come to the exchange rate complication in Chapter 15. Essentially the approach now is partial equilibrium, but this assumption does not limit the applicability of the argument to follow, since it can and will be extended to general equilibrium. The marginal value of extra consumption is equal to the marginal cost of imports when consumption is OB. Hence OB Is the socially desirable level of consumption; it is obtained without intervention. The marginal social cost of production is equal to the marginal cost of imports at output OC. This is greater than actual output in the absence of intervention. Hence intervention, designed to increase output or protect the industry, is required. But this intervention should not, ideally, alter the level of consumption. The aim would be achieved by a subsidy on output at the rate PS per unit, or alternatively, the ad valorem rate PSIOP, It would raise the price received by producers and lead them to raise output to OC. The marginal private cost of production would become CJ. The total cost of the subsidy to the Treasury would be PSJL. Consumers would continue to pay a price of OF for the product. In what sense is this subsidy 'optimum?' And what are the nature and size of the gains that it brings about? Much of the next chapter will be devoted to examining the assumptions implicit in this type of analysis, and then modifying the analysis appropriately. In addition to the two assumptions already listed, four important assumptions are involved, (a) The act of financing the subsidy through taxation does not upset any marginal conditions, for example through making leisure more attractive at the margin than the rewards from effort, and hence reducing incentives, (b) There are no collection costs of taxation. (c) There are no disbursement costs of the subsidy, (d) The redistribution of income from the relevant taxpayers toward the factors of production that produce the protected product and that will gain in income from the subsidy does

By-Product Distortion

9

not represent a net social gain or loss and so can be neglected. From a social point of view, pure income distribution effects either cancel out or are costlessly offset in some way. We now have the following principal result. The social cost of the protected output is the area under the social marginal cost curve, AKLC. The cost of the imports that are replaced is ANLC. Hence the import replacement cost is less than the value of the imports by the shaded area KNL. This is the social gain brought about by the subsidy, A higher subsidy would reduce this total gain since at the margin there would be a social loss indicated by the excess of the social cost curve over the import price OP, A lower subsidy would not fully exploit the potentialities of gain. There are other effects of a redistributive nature, but these are peripheral to the main argument. The relevant taxpayers lose PSJL, which is the cost of the subsidy to them (but not to society). Producers of the product gain PSJN. The beneficiaries of the external economies created by the extra output, whoever they might be, gain KNJL, which is the extent to which social cost of the protected output falls below private cost.

2.2 By-product Distortion Effect of Tariff The same degree of protection could be brought about by a tariff at the rate SPIOP (Figure 2,2). It would raise the domestic price of the product to OS. Production would increase to OC, consumption would fall to OB' and imports would fall to Cff. The production or protection effect would be exactly the same as in the case of a subsidy, yielding a social gain represented, as before, by the shaded area KNL. The difference from the subsidy is that this time there would also be a consumption effect, inflicting a loss. The effort to correct the effects of a marginal divergence on the production side would have as a by-product the creation of a new divergence on the consumption side.

FIG. 2.2

10

Theory of Domestic Divergences

Let us look -At this by-product effect more closely. A tariff that raised the domestic price to OS would reduce consumption to OB'. On the last unit of consumption forgone (represented by point B'), the private and social value, as indicated by the demand curve, is Eff, But the social cost, as indicated by the cost of imports, is only FB*. Hence the conditions for equality of social costs and benefits at the margin required for social optimization would no longer be fulfilled. The consumption of the product concerned which is forgone as a result of a tariff is the amount B'B. If one accepts the simple 'consumers' surplus* approach the value to consumers of this amount is B'EQB while its cost is only B'FQB, so that consumers* surplus of FEQ (shaded in Figure 2.2) is obtained. This would be lost as a result of the tariff. One need not regard the 'consumers' surplus' triangle as measuring the loss precisely, but the basic idea remains that the tariff would impose a new divergence between private and social cost in the process of offsetting an existing divergence. When such a loss is imposed as part of a corrective policy we shall call it a byproduct distortion. This does not mean that it is unimportant. In Figure 2.2, the loss FEQ could well be greater than the gain KNL. It is obvious that a subsidy is preferable to a tariff because the subsidy does not-—at least with our assumptions so far—have an adverse by-product effect. It might also be noted that in the discussion of the tariff we have implicitly made various assumptions similar to those made in the case of the subsidy. Income distribution effects have been neglected. Various effects connected with the use of the customs revenue are ignored. There are assumed to be no collection or administration costs of the tariff. Before going on, let us look at the matter of terminology for a moment. We are using the term marginal divergence to refer to any divergence between marginal private and marginal social costs, or marginal private and marginal social benefits, however caused. If a divergence is caused by government policy of some kind, such as a tariff or other form of tax, then it is a distortion. Thus a distortion is a particular kind of divergence, If a divergence is caused by 'market failure* which is not obviously the result of government policy—for example, a monopoly situation or an externality—then it is not a distortion. We could, if we liked, call it an 'endogenous' divergence, though in general it will just be described as a divergence. To return to distortions, a special type of distortion is the by-product distortion. This is the by-product of a government policy designed to correct, or partially correct, a divergence of some kind, defining the latter in the broadest sense. Many distortions would be of this kind, but one can also envisage distortions that are not the results of corrective policies but are historical relics, are irrational, 'political' in motive, and so on. In the literature on this subject the term distortion has tended to be used in a very broad way to include everything we call here a divergence. We depart here from this broad use of the term distortion in order to narrow its meaning, because it gives the false impression that monopolistic restrictions and

Elaborations

11

externalities are in some sense not endogenous or natural to the economic system and could easily be removed.-^

2,3 Elaborations of the Simple Analysis The partial equilibrium model used is extremely simple, indeed naive. It certainly rests on very many assumptions, some of which will be reviewed in the next chapter, and others later in the book. But nevertheless it is a powerful, and most significant, analysis which can later be generalized, and which will be applied to environmental issues in Chapter 13. Before doing so, it is well worthwhile manipulating it to squeeze out of it several other simple results, Consumption Divergence The marginal divergence may be on the consumption side. Consumption or usage of the product concerned may have external effects, so that, the private and social valuation curves diverge. The product might be gasoline, the use of which pollutes the atmosphere, and hence creates an external diseconomy. The social valuation curve would then be to the left of the private demand curve DD', like RR' in Figure 2.3. Socially optimum consumption would be at OB', If there is no marginal divergence on the production side, production should stay at the nonintervention level QA, First-best policy would be a tax on consumption of the product at the rate PS/OP. This tax should apply to imports and to domestic output at the same rate, so that no protection would be provided.

FIG. 2.3 2 Bhagwati (1971) calls an endogenous distortion what we call an endogenous divergence, a policy-imposed instrumental distortion what we eall a by-product dislortioo, and a policy-imposed autonomous distortion what we just call a distortion that is not a by-product distortion. Essentially the system of classification is the same as Bhagwatrs, The term divergence is used in Meade (195Sa) and was originally used by Pigou (1932).

12

Theory of Domestic Divergences

If a tariff alone were used, the desired decline in consumption could indeed be attained by setting a tariff rate of PSIOP. But a by-product distortion would be created by the protection effect. Domestic output would expand by AC, This extra output would cost ANJCto produce (assuming that private and social cost coincide), but would replace imports that would cost only ANLC. Hence there would be excess cost of domestic production of NJL, this triangle (shaded in Figure 2.3) measuring the cost of the by-product distortion. It is perfectly conceivable that an external economy on the production side coincides with an external diseconomy on the consumption side. It would then be desirable to eliminate both divergences. First-best policy would consist of a production subsidy equal to the externality on the production side combined with a consumption tax equal to the externality on the consumption side. A tariff is the equivalent, from the point of view of the present analysis, of a consumption tax combined with a production subsidy at the same rate. But it would be pure chance if the rate of production subsidy required were the same as the rate of consumption tax required, so it is unlikely that a tariff alone would be & first-best policy. Noneconomlc Objectives The purpose of intervention may be to attain or come closer to some noneconomic objective. It may be desired to expand domestic import-competing production of the product for military reasons, or because some emotional or 'political' value is believed to attach to its production. We need not concern ourselves with the nature of the noneconornic objective, and certainly not with its rationality, but can define it as any objective that comes, essentially, from outside the model. The attainment of such an objective will have an economic cost. There is likely to be, in principle at least, a trade-off between the cost and the noneconomic benefits derived from coming closer to the objective. This can be represented as follows. In Figure 2.4 the private and social marginal cost of production is shown by GG'. The marginal value that society attaches to varying outputs of the product for noneeonomic reasons is shown by the vertical distance between GG' and HH'. As drawn, the marginal social valuation of increments of output from the point of view of the noneeonomic objective first increases and then decreases. At output OC the two curves coincide and the noneconoinic objective has been completely attained. At output OC society regards an extra unit of output from a noneeonomic point of view as worth forgoing JL in real resource cost. Attaining the whole of the noneeonomic objective would be worth to society the whole of the area HGH' between the two curves. HH' is thus, in fact, a special kind of net social marginal cost curve which takes noneeonomic benefits into account.

Elaborations

13

FIG. 2.4

Output is optimum at OC, where the marginal excess economic cost /£, is just equal to the marginal noneconomic value of the output. The net economic cost of expanding output from OA to OC is NJL—th& production-distortion cost of protection. This is the excess cost of protected output over the value of the imports it replaces. This assumes that a subsidy is used to bring about the necessary expansion of output. Noneconomic objectives can be assimilated in our simple analysis. If the noneconomic objective has to do with domestic production, as in our example, a production subsidy or tax will be first-best, A tariff would create a byproduct distortion. If the noneconomic objective is associated with domestic consumption, a consumption tax or subsidy will be first-best. Fixed Targets Sometimes it is convenient to express an objective of policy, whether economic or noneconomic, as & fixed target. There is no question then of trading off the target against the cost of achieving it. In practice no target is likely to be absolutely fixed, quite irrespective of costs. But here, let us assume such a case, since it is an assumption that is often made. The target might be to attain domestic output OC (Figure 2.5). If a production subsidy is used to attain it, an inevitable production-distortion cost of NJL is imposed. The marginal cost of the target is JL. The target could also be attained with a tariff, but then, in addition to the inevitable productiondistortion cost, a by-product consumption-distortion cost would be imposed. Similarly, the minimum cost way of attaining a consumption target would be by a consumption tax or subsidy; again, a tariff would be second-best because it would create a by-product production-distortion cost. The whole of our analysis applies in the case of a fixed target.

14

Theory of Domestic Divergences

FIG. 2,5

2.4 The Second-best Optimum Tariff If there is a marginal divergence in domestic production, social cost being below private cost, a production subsidy is first-best, A tariff inflicts a byproduct distortion. Hence, disregarding any other possible devices, a tariff is second-best. Let us now examine the properties of the second-best solution more closely. We shall suppose that there is a constraint that prevents the use of a production subsidy. Only a tariff can be used. We could think of the constraint as being noneconomic, say political, in nature. Alternatively, we could imagine the administrative costs of a subsidy being so high as to outweigh any possible gain from correcting the production divergence, A tariff could certainly have an adverse effect. The benefit it yields by correcting the marginal divergence in production could be more than offset by the cost of the by-product distortion in consumption. In Figure 2,6 a tariff that completely corrects the production divergence would yield a gain on the

FIG. 2,6

Second-Best Optimum Tariff Divergence in Factor Tariff

15

production side of KNL and a cost on the consumption side of FEQ. The latter cost could clearly exceed the gain. While it is true that a tariff could have an adverse effect, there will be a particular second-best optimum tariff rate that maximizes the gain in the presence of a marginal divergence in production. Provided one makes the simple assumptions implicit in our diagrams, namely, that the demand curve and the supply curve, as well as the social marginal cost curve HH"\ all slope consistently negatively (demand curve) or positively (supply and social cost curves), something fairly precise and interesting can be said about the properties of the second-best optimum solution. Three interrelated properties of this optimum solution should be noted. First, the second-best optimum tariff will be positive but less than the rate of tariff that would correct completely the production divergence. In fact, the correction of the production divergence has to be 'traded off* against the byproduct consumption distortion that the process of making the correction inevitably inflicts (Meade, 1955a, pp. 228-9), In Figure 2,6 the optimum tariff rate is represented by the rate of tariff PP*IOP, How is it arrived at? We can imagine the rate of tariff to be gradually raised from zero. On the first increment in output there is a gain on the production side equal to the whole of the initial marginal divergence, namely NK; as output increases, the marginal gain on the production side gradually declines until, at output OA *, it reaches K*N*. If the tariff had been increased beyond that, output would eventually have reached OC, at which point the marginal gain from a rise in output would have been zero. At the same time the rise in the tariff inflicts a consumption cost. At first the marginal consumption distortion cost is zero; when consumption has fallen as far as OB*, the marginal cost of the last decrease in consumption is F*E*. The net gain consists of the gain on the production side, resulting from the partial correction of the marginal divergence, less the consumption distortion cost. The net gain will be maximized when marginal production gain resulting from an increase in the tariff is equal to marginal consumption cost. In Figure 2.6 this is assumed to be at a tariff rate of PP*IOP, which is thus the secondbest optimum tariff. The shaded area on the left, namely NKK*N*, is the production gain resulting from this optimum tariff, while the shaded area on the right, namely E*F*Q, is the by-product consumption cost. The second-best optimum tariff must be (a) positive and (b) less than PS/OP, which would correct the divergence completely.3 3

In the first edition of this book it was incorrectly claimed that, at the optimum, *K*N* (marginal production gain) is equal to F*E* (margina! production cost).* Britten-Jones ct at. (1987) pointed out thai this would be true only if the slopes of the demand and supply curves were the same. The correct criterion for the optimum is that K*N* \dQ/dp\ equals F*E* |dC/d/»| where the derivatives provide a measure of the marginal change in the quantity produced (Q) or consumed (O in response to a marginal price change due to altering the tariff. They note that the larger the price elasticity of demand and the smaller the price elasticity of supply, the smaller will be the optimal second-best tariff rate.

16

Theory of Domestic Divergences

Second, it is better to impose the second-best optimum tariff than to do nothing at all. There is some rate of tariff that must (given our assumptions) lead to a welfare gain. The size of the gain in Figure 2.6 is the production gain up to output OA*, namely KNN*K* (shaded), minus the consumption distortion cost up to OB*, namely F*E*Q (also shaded). This net gain must be positive, and the second-best optimum tariff maximizes it since at that rate the marginal net gain is zero. Third, the gain from the second-best solution (a tariff) must be less than the gain from the first-best solution (a subsidy). In Figure 2,6, the first-best solution of a fully corrective production subsidy yields a gain, all on the production side, of KNL. The second-best solution falls short of it by the sum of two areas: the production gain forgone by deliberately failing to make a full correction, namely K*N*L, and the consumption distortion cost inflicted, F*E*Q.

2,5 Marginal Divergence in Factor Cost It might be thought that much of the intricate theorizing so far is rather irrelevant. It hinges completely on the importance of the by-product consumption distortion that a tariff inflicts. It might be argued that such consumption distortions are simply not important. This depends, of course, on elasticities of substitution in demand. But, in any case, the real interest in the argument is in the scope it offers for generalization. The simple case where the marginal divergence is associated with production of an import-competing product, and where the only adverse effect of a tariff is the consumption distortion, is heuristieally convenient. Now let us go beyond this case. The marginal divergence between private and social cost might be associated with the use of a factor in a particular industry. For example, the wage that an industry has to pay for its labor may exceed the social opportunity cost of this labor. Other industries are assumed to face no such divergence. This case has provided a popular argument for protection in developing countries. Alternatively, external diseconomies may result from the use of a particular factor or component by an industry, for example, the use of a certain type of polluting fuel. First-best correcting policy then requires a subsidy or tax directly related to the relevant factor input—a wage (or employment) subsidy to the industry in the first case, and a tax on the use of the fuel by the industry in the second. Let us take the first case as our example. The wage subsidy will have two favorable effects. First, labor will be substituted for other factors in the industry to an appropriate extent. This assumes that there are not technically fixed proportions between them. Second, the costs of the industry will fall, and hence its output will expand relative to other industries. A production subsidy to the industry would be second-best. It represents, in effect, a subsidy to all factors used by the industry, both labor and nonlabor, and hence would lead to the by-product distortion of overuse of nonlabor

General Equilibrium

17

factors in the industry. It would fail to make the industry's production more labor-intensive. Even the first-best solution may require some inflow of nonlabor factors into the industry, though proportionately less than the inflow of labor (unless there are fixed proportions). But the production subsidy would lead to a greater inflow of nonlabor factors for any given expansion of the industry's use of labor. The second-best optimum production subsidy will result from trading off the gains from offsetting the marginal divergence in labor cost against the byproduct distortion, along the lines of our earlier analysis of the second-best optimum tariff. It will normally involve a lesser stimulus to the industry's use of labor than the first-best policy would have done; it will lead to some welfare gain, but this gain will be less than the gain from the first-best wage subsidy. A tariff is third-best, since it imposes not only the same by-product factor distortion that the production subsidy imposes, but in addition inflicts a byproduct consumption distortion. There will be a third-best optimum tariff. It will normally involve a lesser stimulus to the industry's use of labor and to its output than the second-best optimum production subsidy would have done, and it will lead to some welfare improvement, but less than the second-best optimum policy. It might be noted that one could interpret the concept of a marginal divergence so widely as to incorporate all possible considerations which call for some kind of social intervention, not just the ones we have mentioned so far. There might be monopolistic pricing. There might be a distorting tax which, for one reason or another, must be taken as given. The divergence may be, not between private and social interest, but bet ween privately perceived private interest and socially perceived private interest, the latter based perhaps on more mature judgment or better information (as in the case of smoking or drugs).

2.6 General Equilibrium, and Many Commodities The method of exposition so far has been partial equilibrium. But this has only been a matter of exposition. The analysis is fully applicable in general equilibrium. Indeed, if one stays within the constraints of a two-commodity, twofactor general equilibrium model of the type generally used in international trade theory, the translation to general equilibrium is easily made. There is a large literature which expounds, with much elaboration and sophistication, many of the arguments advanced in this chapter in terms of two-good twofactor geometry. A two-sector general equilibrium exposition of the main arguments is given in Appendix 2.1. Even our partial equilibrium diagrams can be translated into general equilibrium. The product can be defined as the economy's sole importable, and price and marginal cost can be redefined respectively as the relative price ratio in a two-good model and the marginal cost of the importable in terms of the

18

Theory of Domestic Divergences

exportable. Similarly, the demand curve can be interpreted as showing the marginal value to consumers of the importable in terms of the exportable. One would have to be careful about using the consumers' surplus method (for which purpose one would need to use a constant utility demand curve, hence excluding income effects), but the main ideas would stand, In the real world of many commodities and many factors, everything becomes, of course, much more complicated. One could describe numerous possible complex cases, with a variety of interacting divergences, to show what, in each case, the appropriate first-best package of policies would be, and how some other package would compare with it. The optimal correction required for one particular divergence may well depend on whether another divergence has been corrected. If a second-best policy is used to correct one divergence, hence creating a byproduct distortion, it may create the need for a supplementary policy designed to correct, at least partially, the newly created distortion. In general, one can say that the first-best package of corrective policies will consist of policies that get as close as possible to the sources of the various marginal divergences, and hence that minimize the by-product distortions that are created. The case where it is desired to foster domestic production of an importcompeting intermediate good is rather interesting, and illustrates some of the issues. External economies may be believed to attach to domestic production of an import-competing intermediate product A which is an input into another import-competing product, B. Let us assume that this externality is initially the only marginal divergence and that the only policy instruments available are tariffs. Hence a tariff is imposed on A. This raises the costs of producers of B, imposing negative effective protection on them, so leading to the by-product distortion of underproduction of B. A second corrective policy, consisting of a tariff for B, needs then to be imposed. The second-best optimum tariff rate will not be so high as to restore the effective rate of B to zero since now the tariff will impose a by-product consumption distortion. But finally there will be an optimum set of a tariff on A and a partially compensating tariff on B, the whole exercise being designed to correct the marginal divergence in A, taking account of the various by-product effects.

2.7 Home-market Bias A country may foster the development of its manufacturing industry with tariffs or import restrictions and create a large and sophisticated industrial sector, one which might be expected to export a significant part of its output. But if the protection is only provided for the home market, with little special encouragement for exporting, lopsided development, with very few exports of manufactures, may result. This indeed, was the result of the policy of

Home-Market Bias

19

'import-substituting industrialization' which was pursued from the 1950s to the 1970s in many countries, notably Argentina, Brazil, and Mexico, and also in India and Australia. It came to be gradually realized that one of the main weaknesses of the so-called import substitution strategy is home-market bias: its failure to encourage exports of manufactures. Hence opinion swung in favor of measures which in some form or other subsidized or encouraged these exports.4 It is of course assumed that there is some general case for fostering manufacturing, presumably at the expense of agriculture and mining, A Three-product Model The nature of home-market bias can be analyzed rigorously and related to our general analysis. It is an important by-product distortion created by a tariff system, one not brought out in our simple partial equilibrium analysis earlier. Let us assume that three products are produced and consumed in the country. Two products are exported as well as consumed domestically, namely Xa, the agricultural product and Xm, a manufactured product. Another manufactured product, Mm, is imported and also produced domestically. Now we make two assumptions. First, external economies attach to production of both types of manufacturing, so that their marginal social costs of production fall below their private costs by a given percentage, the same for both of them. 'Cost' is defined in terms of agricultural production forgone. In the absence of any intervention, production of Xa will be too high from a social point of view and of Xm and Mm too low. Second, relative price changes over the relevant range do not induce any consumption substitution between the three goods. This assumption is to isolate production effects; it will be removed later. Subsidy and Tariff Compared The first assumption means that optimum output could be attained by a subsidy to production of manufactures equal to the rate of external economies, financed by a nondistorting tax. Alternatively, agricultural production could be taxed, the revenue being redistributed in a nondistorting way. Essentially, both methods are equivalent to a tax on agriculture which finances a subsidy on manufacturing. This subsidy to manufacturing would bring about a resource movement out of agriculture into the two manufacturing industries as indicated by the arrows in Figure 2.7. Exports of Xa would fall, exports of Xm would increase, and imports of Mm would fall.

FIG. 2.7 4

See Little et ai. (1970), especially pp. 128-34.

20

Theory of Domestic Divergences

If, instead of the production subsidy, a tariff were imposed, the domestic price of Mm would rise relative to the domestic prices of the two export products, and resources would be drawn into Mm out of both export products. The resource movements induced by the tariff are shown by the arrows in Figure 2,8,

FIG. 2.8

When we compare the tariff with the production subsidy, we notice two differences; first, the production subsidy leads to a movement of resources out of agricultural exporting into manufactured exports, while the tariff fails to bring about such a shift in the pattern of exports; second, the tariff leads to a switch of resources within the manufacturing sector, out of manufactures for exports into import-competing manufactures. If the rate of tariff is just sufficient to increase total manufacturing production to the same extent as the optimum rate of production subsidy would have done, the movement of resources out of Xa will be the same in both cases. The difference between the two cases will be in the pattern of manufacturing production. The tariff will lead to an expansion of import-competing manufacturing production Mm and a contraction of Xm, though production of the two combined (valued at world prices) will have expanded; the production subsidy will lead to a lesser expansion of Mm and in addition to an expansion of Xm. Protection versus Promotion Home-market bias would be removed if the tariff on Mm were supplemented by an export subsidy for the manufactured export. (To avoid reimport of the subsidized export, Xm, a tariff may also need to be put on imports of Xm.) If the rale of export subsidy were the same as the rate of tariff for the importcompeting manufacture, Mm> the domestic prices of both manufactured goods would rise to the same extent, and the only resource movements would be out of Xa into both manufactures. When a tariff alone is imposed, manufacturing is only protected; when the tariff is supplemented by an export subsidy for manufacturing, then manufacturing is promoted.^ Protection creates a homemarket bias; promotion avoids it. The method of combining a tariff and an export subsidy at the same rate would have exactly the same effect as a production subsidy for manufacturing (given that we are assuming no consumption effects). Hence there are two methods of promotion. Both raise the prices received by manufacturers relative to the prices received by agricultural producers, and so bring about the desired resource movement. It is possible that in the absence of any intervention the country does not export any significant quantities of manufactured products. Hence, in terms of 5

The distinction between 'protection' and 'promotion' comes from Little et al. (1970).

Hierarchy of Policies

21

our formal model, initially Xm is an imported product. But when the optimum subsidy is imposed, its production would expand so much that it would begin to be exported. Thus optimal policy would require trade reversal with respect to this product. A tariff alone could not bring about this trade reversal. At the most, it could bring about complete replacement of imports of the product by domestic production,6 But any increase in the tariff beyond the point where imports have ceased would have no effect. To bring about the optimum output of Xm and hence the necessary trade reversal it would be necessary to supplement the tariff with an export subsidy. Otherwise there will, as before, be underexpansion of Xm.

Consumption Effects We could now introduce the familiar consumption effects. Both tariffs and export subsidies will raise prices to the domestic consumers of the protected products and so bring about an undesired switch in the consumption pattern, Our main argument that promotion is preferable to protection is not affected. But it does mean that the two forms of promotion—tariff-cwn-export subsidy on the one hand, and production subsidy to manufacturing on the other—no longer have identical effects. The production subsidy does not create a consumption distortion and is thus clearly preferable. It could also be shown that, if one must use a tariff-cwm-export subsidy as promotion, rather than using a production subsidy, the rate of export subsidy should not necessarily be identical with the rate of tariff. The aim of nonuniformity would be to minimize the inevitable consumption cost.

2.8 The Hierarchy of Policies For any given marginal divergence, or set of divergences, there is a first-best optimal policy or set of policies. Essentially this policy involves making the appropriate correction as close as possible to the point of the divergence. But many policies may be conceivable, and one should be able to order them in a hierarchy of policies, from first-best to second-best, and so on. This is an interesting exercise, for it brings out the logic of the general approach, and especially the effects of imposing constraints on the choice of policies which compel a movement down the hierarchy. At each step down the hierarchy an additional by-product distortion is imposed, and the welfare level attainable with the appropriate optimal policy * This is definitely true only in the formal model used here, with an upward-sloping supply curve and perfect competition. If the industry were monopolized and subject to economies of scale, it might become possible that in the absence of a tariff a product would be wholly imported, while the tariff caused domestic producers to supply the whole domestic market and also to export. See section 7.7,

22

Theory of Domestic Divergences

declines. Furthermore, given the appropriate optimal policy, the extent of the correction to the basic divergence will normally decline. We shall now look at a particular case of a marginal divergence to illustrate this approach. But it must be stressed that the methodology is perfectly general. It can be applied to the analysis of economic policies in all spheres, not just trade policies. For any divergence between private and social costs or benefits, there is a hierarchy of policies, It is always true that as one goes down the hierarchy an additional by-product distortion is imposed, the welfare level attainable declines, and the extent of the optimal correction to the original divergence will normally decline. This approach will be used continually throughout this book, and perhaps it should be used much more widely when economists describe or analyze alternative economic policies. For example, it could be used in comparing various policies to deal with external diseconomies, such as pollution or traffic congestion. The approach will be qualified or elaborated somewhat in the next chapter. Suppose that the marginal private cost of labor to the manufacturing sector in a developing country exceeds the social opportunity cost of this labor. Assume that this is the only significant marginal divergence. Figure 2.9 represents the relevant hierarchy of policies. First-best policy is to subsidize the use of labor in manufacturing. (Alternatively, if it were possible, labor in agriculture could be taxed.) Second-best policy is to subsidize manufacturing production; this, as already pointed out, will impose the by-product distortion of overencouraging the use of non-labor factors (notably capital) in manufacturing. Third-best policy is to impose a tariff on imports of manufactures combined with an export subsidy for manufactures at the same rate. This will add a by-product consumption distortion. Fourth-best policy is to impose a tariff on manufactured imports on its own. This will add a third by-product distortion, namely home-market bias. The tariff thus creates three by-product distortions. The fourth-best optimal tariff may need to be set at a level that will yield a stimulus to the employment of labor in manufacturing that is much less than the stimulus the first-best wage-subsidy policy would have provided. The gain from correcting the basic divergence has to be traded off against no less than three by-product distortions. There is an element of arbitrariness in the choice of policies to include in a hierarchy of policies. One could include all technically feasible policies. Alternatively, one could limit oneself to those policies which have been under consideration, or perhaps which have actually been used in the past, or which are politically conceivable. The example just given has not been comprehensive; the aim has been to display the special features of the tariff as a policy instrument. This is brought out in Figure 2.9, which shows two other policies that might be included in the hierarchy, A subsidy on production of import-competing manufactures imposes a home-market bias but no consumption distortion. In the hierarchy it is thus

23

Trade Divergences THE HIERARCHY OF POLICIES PfiV8t& GQSt Of iS/faOU!1 t&

manafscfuflnQ &xc$&ct$ the soda/ opportunity cost

By-product distortions

(1) Subsidy to labour CO$1

in ftusnyfactuHng

None

(2) Subsidy to manufacturing production

(3a) Tariff plus •Xpert subsidy fof m*nuf*ctur«*

Labour-intensity too low

1. Labour-intensity too low 2. Consumption distortion

(3b) Subsidy to importcompeting manufacturing production

1. Labour-intensity too low 2. Home-market bias

(4»)

1. Labour-intensity too low 2. Consumption distortion 3. Home-market bias

T»r*«

(4b) Export subsidy

1. Labour intensity too low 2. Consumption distortion 3. Pro-trade bias

FIG. 2.9

graded as third-best, alongside the tariff-c«m-export subsidy. Similarly, an export subsidy on manufactures on its own parallels a tariff on its own, both being fourth-best; the subsidy creates pro-trade bias and the tariff home-market bias. When two policies are third-best in the hierarchy we are simply saying that they both add one more by-product distortion to the second-best policy. One of the 'third-best' policies may lead to a higher welfare level than the other, so that one would have to renumber the policies if one wished to order them strictly on the basis of welfare levels attainable, and not just on the basis of the numbers of by-product distortions imposed.

2,9 Trade Divergences So far we have assumed that all divergences are domestic divergences (and we shall maintain this assumption in the next three chapters). Thus in our various

24

Theory of Domestic Divergences

diagrams we have assumed that the import price OP indicates both marginal private and marginal social cost of imports. But it is possible that there is a trade divergence. In that case a trade intervention will indeed be first-best. The most important case is where the country can affect its terms of trade. We shall discuss this at length in Chapter 6, and here we continue to make the small country assumption that the country faces given import and export prices at any point in time. But there may be other reasons why private and social costs of importing (or returns from exporting) diverge. Some kind of external diseconomy may attach to trade as such. If this is so, in our simple model, a tariff will be a first-best policy. It will be optimum to restrict trade by both increasing import-competing production and reducing consumption of imports. In Figures .2.1 to 23, if the marginal social cost of imports is OS, then a tariff of PS/OP will bring about optimum domestic production and consumption. It is consistent with our general approach that when a marginal divergence is genuinely associated with trade as such, the first-best policy is a trade intervention. There might be a noneconomic objective connected with trade, namely to make the country more self-sufficient, perhaps for nationalistic reasons. It is only by positing such an objective that one can make sense of some seemingly irrational policies. In terms of our model, the aim is then to reduce imports. This will inevitably be attained at a cost -the sum of the production and consumption distortion costs of protection. A tariff will then be first-best. Since the country is indifferent to the way in which imports are reduced—whether it is by increasing domestic production or reducing domestic consumption—the reduction should be brought about so as to equate marginal productiondistortion cost to marginal consumption-distortion cost: this will minimize the total cost of achieving any given degree of import restriction, and will indeed be brought about by a tariff. One could suppose that some marginal value is attached to the nationalistic, objective, perhaps declining as the volume of trade is reduced. This marginal value must then be equated to the marginal cost of attaining the objective, and this can be brought about by fixing the tariff equal to the marginal value of the objective. A frequent misunderstanding should be avoided. The foreign suppliers of the product may be subsidized or taxed in some way so that the costs of their exports—our country's imports—are not equal to marginal social costs in the supplying country. Alternatively, there may be some externality in the foreign country. But this does not mean that there is a marginal divergence from our country's point of view. If the price is given to our country, then, unless some particular advantage or disadvantage attaches to foreign trade in the product as such, there is no marginal divergence on the trade side. Our country can ignore the marginal divergence abroad. This approach assumes that it wishes to apply a national, and not a cosmopolitan, welfare criterion, Production of the product in the foreign supplying country may pollute the latter's environment

By-Product Distortion

25

and so inflict external diseconomies. This may well present an argument for social intervention in the foreign country. But from the purely selfish point of view of the importing country, there is no argument for intervention. It is particularly important to note—because it is often not understood— that a foreign country may be subsidizing its exports, and this may well create a marginal distortion abroad, but it does not mean that there is a marginal divergence for our country which is importing the product. If the import price is given, then it is indeed given, and it is not really relevant whether it has been brought down by the improved efficiency of foreign producers, a decline in foreign demand for the product, or foreign subsidies. It follows that a free trade policy may be optimal for our country even when its trading partners are intervening in trade. The central issue is whether the import price facing us is indeed given. It is possible that our country's tariff may cause the foreign country's intervention policy to change; but this is another matter to which we turn in Chapter 6. Finally, there is the question of the exchange rate. Let us again make the small country assumption and assume that there are no domestic divergences of any kind, and no trade divergences except one. This is that the country is running a balance-of-payments deficit which has to be ended or reduced. We can think of this most conveniently as a fixed target. Hence foreign exchange is wrongly priced, so that there is a trade divergence. First-best policy is to raise the price of foreign exchange appropriately (devalue the domestic currency). This will cause the pattern of domestic production to shift from nontraded to traded goods and the pattern of consumption in the other direction. The use of consumption or production-switching measures alone (such as subsidies on import-competing production) would be second-best or worse. It is not too difficult to prove this rigorously, although we shall not do so here. Since the divergence is in the price of foreign exchange, the use of a tariff or a set of tariffs will not be first-best, A tariff will only act on the import side, but will fail to stimulate exports. Hence trade intervention that uniformly raises the domestic prices of imports and exports is required. This can be attained by altering the exchange rate itself or by combining a uniform ad valorem tariff with a uniform ad valorem export subsidy at the same rate, tariff and subsidy applying to all trade, visible and invisible.

APPENDIX 2.1: The Theory of Domestic Divergences in Two-product Geometry It is possible to represent some of the main ideas of the theory of domestic divergences by using the two-product geometric method familiar to all students of international trade theory. The method is elegant. It is also attractive that the approach is explicitly and clearly general equilibrium. Hence it is comforting to find that the results obtained from the simple demand-and-supply diagrams used earlier are confirmed here. First we represent the central ideas of sections 2,1,2,2, and 2,5, At the end a special case, where trade reversal would be the first-best optimum, is analyzed. In Figure 2,10 the importable, product M, is shown along the vertical axis, and the exportable, product X, along the horizontal axis. The quadrant contains a map of community indifference curves which represent ordinal welfare levels as well as community preferences on the assumption that a given income distribution is maintained by an independent policy- that is, that a social welfare function is being consistently and costlessly applied.

FIG.2.10 The social production transformation curve is HH'. It shows full employment production possibilities. Its slope at any point indicates the social marginal rale of transformation at that point. The privately perceived transformation rate

Two-Product Geometry

27

differs from this. Thus at point P the latter is assumed to be given by the slope of SS* and at P' by the slope of If'. These private transformation rates are flatter than the social ones; hence we are assuming that the marginal private cost of transforming X into M exceeds the social cost. This is the case that was analyzed in section 2.1; an external economy is associated with production of the importable. It should be noted that in this case the economy will be producing on its production-efficiency frontier. In terms of the EdgeworthBowley box diagram, it is producing on the contract curve. The case where it is not producing on this frontier will be considered later. Free Trade Equilibrium The slope of SS' is assumed to indicate the given world price ratio. In free trade, output is at P, and the country can trade along SPS', choosing to consume at C, on the income-consumption line OZ appropriate to the world price ratio. At Can indifference curve is tangential to SS'. The rate at which Xcan be transformed into M through foreign trade is the foreign marginal rate of transformation or FRT. The slope of the indifference curve is the domestic marginal rate of substitution in consumption or DRS. The slope of the domestic production transformation curve Htt at the relevant point gives the domestic marginal social rate of transformation in production, or DRT. In the free trade situation the foreign marginal rate of transformation is equal to the domestic marginal rate of substitution in consumption, but both differ from the domestic marginal social rate of transformation in production (i.e. FRT ~ DRS # DRT), The Optimum In the first-best optimum situation the three marginal rates would all be equal. This is attained by a production subsidy on M (or a production tax on X) which shifts output to P', the rate of subsidy being given by the excess of the private transformation rate at P' (given by the slope of tt') over the social rate (slope of HtF), It is, of course, assumed that the subsidy is financed by costless nondistorting means. At point P the country can trade along JP'J' (parallel to SPS') and will choose consumption point E. At this point an indifference curve is tangential to JP'J'. The point £"is on the same income-consumption line OZ as the free trade consumption point C. It is the best point attainable with the given production transformation curve HH' and the given world price ratio.

A Tariff Now consider a tariff (or export tax). We refer to Figure 2.11. Imagine the tariff to be raised from zero. The output point will move along HH' from P toward F and perhaps beyond. This is the production effect of the tariff. For each production point the trading possibilities are given by a line through that

28

Theory of Domestic Divergences

FIG, 2,11 point having the slope of the world price ratio. Thus, when production is at P' the trading possibilities are along//3'/'. At the same time, as the tariff is raised, the relative price of importables facing consumers increases above the free trade price ratio, so that the relevant income-consumption line swings to the right. We are assuming that tariff revenue is returned to consumers, and that all of it is spent in accordance with the demand pattern indicated by the community indifference curve map. If a tariff were imposed that was sufficient to eliminate the production divergence and hence to bring production to P'—the first-best optimum output—the consumption pattern would move to the income-consumption line OZ', and the consumption point would be C. It can be seen that the indifference curve through C could be higher or lower than the curve through C (which was the free trade consumption point); in other words, if a tariff is used to eliminate the production divergence completely, there could be a gain OF a loss relative to free trade.

Two-Product Geometry

29

The Meade Curve and the Second-best Optimum There are many possible tariff levels and hence consumption points. Suppose a tariff shifts output to P". The country can then trade along GP"G', With the price facing consumers above the free trade price, they will choose to consume at C", on the relevant income-consumption line OZ". The slope of the indifference curve at C" is flatter than at C, and is the same as the privately perceived marginal rate of transformation in production at P" (indicated by the slope of the line qq' through P"). One can thus trace out a curve CC" C to he called the Meade curve—which shows the locus of consumption points as the tariff is raised from zero toward the rate at which the production divergence is eliminated.7 The tariff could be raised further, so that this curve could be continued to /?', the self-sufficiency point, (The curve could also be carried to the left of C, the range RC resulting from an import subsidy.) Somewhere on this Meade curve CK an indifference curve must be tangential to it; this will indicate the best consumption that can be obtained with a tariff. It would result from the second-best optimum tariff. In Figure 2,11 this second-best optimum consumption point is C", with the associated production point P", The optimum point on the Meade curve must be between C(the point of free trade consumption) and C (the consumption point if the production divergence were completely eliminated). The production effect of a marginally positive tariff will bring the consumption point to a trading possibility line above SS', in fact extending consumption possibilities outside the free trade consumption possibility line SS'. Furthermore, the curve must be tangential to //' at C, and hence at C an indifference curve must cross it (since the slope of the indifference curve at C' is equal to the slope of the privately perceived transformation rate at P'), It follows that the second-best optimum tariff is positive and should be less than needed to bring production to P' where the whole of the production divergence would be eliminated. In this second-best optimum situation DRS > FRT > DRT, and hence the equality DRS = FRT which existed in free trade has been destroyed. It is evident from Figure 2.11 that the indifference curve through C" must be lower than that through E. In other words, the second-best optimum policy attains a lower welfare level than the first-best (production subsidy) policy. Trade Reversal The first-best optimum solution could call for a production subsidy on M which is sufficient to reverse the pattern of trade and turn M into the exported good. This case is represented in Figure 2.12. Free trade production is at P and ^ It is termed the Meade curve here because it is used to represent ideas that originated in Meade (1955a) (even though the curve itself cannot actually be found in any of Meade's books).

30

Theory of Domestic Divergences

consumption at C. First-best optimum production is at F, as before, but the optimal consumption point, with given production at F, is at E to the right of F. Hence in the optimal situation production of M would exceed consumption, and it would be exported. What is the second-best optimum in this case? Imagine the tariff to be raised from zero until trade is eliminated. The resultant consumption points would trace out the Meade curve CR', derived as in Figure 2.11. A tariff on its own cannot carry consumption to the right of R', since once production and consumption have reached that point any increase in the tariff would be redundant. The optimum will now be at the point where an indifference curve is tangential to CR', or if none is, at the self-sufficiency point R' itself. A tariff on its own cannot, of course, bring about trade reversal. The most it can do is to eliminate trade. But it may not be optimal to eliminate trade even though the first-best optimum requires trade reversal. Suppose we allow for both a tariff and an export subsidy as instruments of policy. The Meade curve can then be continued beyond R' to C, and even beyond that. Thus a tariff combined with an export subsidy (both at the same rate) could bring production to F, and hence consumption to C. Any point to the right of R' represents trade reversal. It is interesting to note that in this case it may not be second-best optimal to bring about trade reversal even though trade reversal could be achieved and the first-best optimum requires trade

Two-Product Geometry

31

reversal. This would be so if an indifference curve were tangential to the extended Meade curve CR'C to the left of R'. Some imports of M would then continue. Factor Market Divergence The domestic divergence might be in the factor market rather than associated with production of a product as such. An extensive literature has analyzed the case of a fixed wage differential in the two-sector, two-factor model, one sector paying a higher real wage than the other,8 In the absence of intervention the economy will then produce within the social production transformation curve, The marginal rate of substitution of one factor for another along an isoquant will not be the same in both sectors; in the Edgeworth-Bowley box diagram, the dimensions of which show the country's fixed factor stocks, the country will not be producing on the contract curve which traces out points of tangency of the isoquants. Hence a Paretoefficiency condition will not be fulfilled. There will be a distorted or 'shrunk-in' production-possibility curve within HH' (Figure 2.10) which traces out possible production points given the wage differential, the actual production point being determined by the product price ratio. (The curve is not drawn in Figure 2.10.) A first-best policy of appropriate wage subsidy will bring the economy to the first-best solution represented in Figure 2.10 (production at P: consumption at £), But this time a production subsidy to the industry paying the excessive wage will be only second-best. The second-best optimum production subsidy will bring the economy to the point where the given world price line is tangential to the 'shrunk-in* curve. ° See Hagen (1958), Bhagwati and Ramaswanii (1963), and the extensive subsequent literature surveyed in Magee (1973),

APPENDIX 2.2: Theory of Domestic Divergences: Notes on the Literature The theory of domestic divergences originated in Chapter 14 of Meade (1955a), though this has often been forgotten. The chapter is entitled 'The Second-best Argument for Trade Control: (3) Domestic Divergences.* The important proposition that a subsidy is better than a tariff to deal with a marginal divergence is made clearly on pp. 230-1 of that book, Meade also expounded the concept of the second-best optimum tariff (pp. 228 30), Furthermore, he pointed out that a factor subsidy is the best device to use when there is a factormarket divergence (pp. 232-4), The modern rigorous analysis of factor-market distortions can be said to have begun with Hagen (1958), who noted that a subsidy per unit of labor, and not a tariff, is first-best to offset a wage-rate divergence (owing to a wage differential), An influential article by Bhagwati and Ramaswami (1963) stressed the main point about the optimality of dealing directly with a 'distortion,* and the crucial relevance of the relationships between domestic rate of substitution (DRS), domestic rate of transformation (DRT), and foreign rate of transformation (FRT), This was followed by the important article that really focused on the central issue, applied it to various cases, and set one of the main themes of the present book, namely Johnson (1965). For the early history one should note three other publications: Pigou (1932) introduced the concept of divergences between marginal private and social costs and benefits, these divergences then calling for 'Pigovian* taxes and subsidies. Haberler (1950) analyzed systematically for the first time the gains from trade (no trade compared with free trade) in the presence of various domestic divergences; this article influenced Bhagwati and Ramaswami (1963) and Johnson (1965) in their choice of problems, but was not concerned with the question of the optimal policy instrument. Furthermore, Corden (1957b) showed geometrically that in the small-country case a subsidy is preferable to a tariff, if a given amount of protected production is desired, and also drew attention to the DRS, DRT, FfiTdistinetions. Many other articles, mainly by Bhagwati, Ramaswami, and Srinivasan, in various combinations, have been published on the subject. These have been consolidated in Bhagwati (1971), which is now the most frequently cited paper on the theory of domestic distortions.

3

The Four Assumptions of the Theory of Domestic Divergences The whole argument of the previous chapter has involved four assumptions, ( I ) Subsidies can be financed by 'nondistor ting' taxes, (2) Taxation involves no collection costs. (3) There are no costs of disbursement of subsidies, (4) The income distribution effects of various policies (such as the redistribution from taxpayers to subsidy recipients) can be neglected. These four assumptions are generally made in the numerous articles that have been published advancing the 'theory of domestic distortions.'1 The question is whether the central arguments of this approach still stand when the assumptions are removed, especially the argument that taxes and subsidies on trade are never first-best policies when the divergences between social and private cost or benefit are domestic in nature, The issues involved are obviously important. If the basic approach really depended on these assumptions, then it could be rather easily dismissed as academic- as indeed it has been dismissed by many a practical-minded reader. It is rather surprising that, in spite of the surfeit of articles in the economic journals on the 'theory of domestic distortions,* no one other than Meade (I955a) has found it necessary to remove or review these assumptions, or, in the case of assumption (3), even to state it, Meade removed assumption (1) and brought out some of the considerations discussed in sections 3, 1 and 3,2 below. Each of the assumptions will be considered here in turn. It will be shown that the analysis certainly needs to be modified, but the central argument is unshaken by the removal of the important assumptions (I) and (4). Removal of assumption (2) slightly dents it, while removal of assumption (3) does affect it. ' The two standard references are surprisingly strongly worded on this subject. The contention that the payment of subsidies would involve the collection of taxes which in practice cannot be levied in a non-distortionary fashion is fallacious. A tax-c «»i-subsidy scheme could always be devised that would both eliminate the estimated divergence and collect taxes sufficient to pay the subsidies' (Bhagwati and Ramaswami, 1963, p. 50). '. . . it is assumed in this paper, in accordance with the conventions of theoretical analysis of these problems, that government intervention is a costless operation: in other words, there is no cost attached to the choice between a tax and a subsidy. This assumption ignores the empirical consideration, frequently introduced into arguments about protection, that poor countries have considerably greater difficulty in levying taxes to finance subsidies than they have in levying tariff's on imports. This consideration is of practical rather than theoretical consequence, and to constitute a case for tariffs requires supplementation by empirical measurement of both the relative administrative costs and the economic effects of promoting favored industries . . . " (Johnson, 1 965, p. 7).

34

Four Assumptions of Theory

We begin with removing assumption (1). At the end some doubts about the whole 'subsidy-biased' approach will be introduced.

3.1 By-product Distortion Costs of Revenue-raising: Subsidies Remain First-best There is no practical way of collecting taxes that does not impose costs. Income taxes distort the choice between work and leisure. Taxes on commodities at various rates distort the choices between these goods as well as between the taxed goods and leisure. These used to be called the 'deadweight'costs of taxation and result from production and consumption distortions created by taxation. In addition there are collection costs, which we shall consider more fully shortly. Economists sometimes talk about 'nondistorting poll taxes." This concept is a useful theoretical fiction but no more than that. Raising revenue imposes inevitable by-product costs the 'deadweight* distortion costs and the collection costs. It was the theme of the previous chapter that when a domestic divergence is corrected by a subsidy at the point of the divergence, no by-product distortion is imposed. By contrast, a tariff imposed a by-product distortion, or possibly even several kinds of them. On this basis it was argued that the subsidy was first-best. Now we find that the process of raising revenue to finance correcting subsidies also imposes inevitable by-product costs. Has the whole argument then been destroyed? Can one argue that, since each method imposes costs, there is no a priori way of deciding which method will make a given correction at minimum cost? Surprisingly, such an argument would be fallacious. Let us assume for the moment that there are no collection costs of taxation, including tariffs, and no disbursement costs of subsidies. The correct approach appears then to be the following. The first-best way of correcting the consequences of an excess of private over social cost to a given extent is to provide a subsidy and then to finance this subsidy by a minimum-cost package of taxes. Alternatively, first-best policy may require a tax at the point of the divergence, with the revenue raised being remitted in a minimum-cost way. To return to the case where a subsidy is required and has to be financed, the minimum-cost tax package is one that raises a given amount of revenue in a way that minimizes the by-product distortion costs of doing so. By-product costs or distortions cannot be avoided, but they can be minimized. The considerations influencing the composition of the minimum-cost tax package will be discussed in more detail in the next chapter. This first-best approach of protecting an industry with a subsidy and then financing the subsidy in a minimum-cost way can be compared with the secondbest method of providing protection for a particular product wholly by a tariff on that product. It is then important to appreciate the characteristics of a

Subsidies Remain First-Best

35

tariff. It is the equivalent of a tax on consumers of the product concerned, the revenue from which finances a subsidy to the domestic producers of this product, the rates of consumption tax and production subsidy being the same. In addition, if some imports remain, the tariff would yield customs revenue; it would then be equivalent to a tax on consumers that is greater than is needed to finance the subsidy.2 Hence the use of a tariff on its own in order to protect the product means that the subsidy is financed in a very particular way—solely by a tax on consumers of that particular product, the rate of tax being the same as the desired rate of subsidy. By contrast, the first-best method of protection involves no such constraint: the subsidy is simply financed by a minimum cost tax package, It is most unlikely that the minimum cost way of raising given revenue would be solely by a consumption tax or tariff on this particular product at the rate of the desired subsidy. There may be far less distorting ways of raising the same revenue (Meade, 1955a, p. 231). Our main point can be made by considering an extreme case, Suppose that the country cannot or will not use any taxes other than tariffs. Thus the subsidy would have to be financed from customs revenue. Suppose, further, that there are only two importable products, X and Y. There is, as before, a marginal divergence in production of X, requiring it to be protected. Even then, a tariff on X to achieve the whole of the desired protection would probably not be firstbest. First-best policy would be a subsidy on production of Jf financed by that combination of tariffs on Jand Y that inflicts minimum distortion costs. Because there will be some tariff on X, the rate of subsidy required will indeed be less than in the absence of such a tariff, but the main point is that a tariff on the other product is likely to be part of the minimum-cost package.

3.2 Collection Costs Let us now remove assumption (2) and allow for collection costs of taxes. These are not only the costs of tax administration to the Ministry of Finance but also the costs imposed on taxpayers in their efforts to comply with the tax law while minimizing their tax burdens, or to evade it. The by-product costs of taxation thus consist of distortion costs and of collection costs. The interesting question is whether collection costs alter the argument that domestic divergences should be corrected by subsidies and taxes close to the points of the divergences, rather than by trade policies. Suppose initially that collection costs of tariffs are low or zero while there are significant collection costs for other types of taxes. It will then remain optimal to subsidize, and to finance the subsidy with a minimum-cost tax package. This * This could be illustrated with Figure 2.2. The lax on consumers per unit is PS. This consumer tax raises revenue of PSEF, This revenue finances the production subsidy of PSJL and raises customs revenue of LJEF.

36

Four Assumptions of Theory

package must take collection costs into account and for this reason may well consist mainly of trade taxes. It will be optimal to allow some increase in distortion costs in order to keep down collection costs. For the limiting case where the tax package would consist wholly of trade taxes, the argument in favor of subsidizing and then using several tariffs to finance the subsidy rather than simply imposing a protective tariff, has been made already. This argument underlines that while high collection costs for nontrade taxes may make the use of trade taxes first-best for raising revenue, these costs do not justify the direct use of trade taxes to correct domestic divergences. Once we allow for collection costs on trade taxes, the argument may have to be qualified, though this qualification is probably not of great importance. Suppose, as before, that product X is to be protected and that any subsidy would have to be financed by revenue tariffs on Xand Y. If there are high collection costs for the tariff on Y, it may be more economical to achieve the whole protection with a tariff on, X, where collection costs may be low because imports will be low. Since distortion costs will exceed those that would result if X were subsidized and financed by tariffs on I'and Y, the extra distortion costs must be set against the lower collection costs. The general argument is clearest when the optimum tariff on X would be a prohibitive one. Collection (or administration) costs would probably be very low, much lower than when nonprohibitive tariffs on X and Y have to be imposed to finance a subsidy for domestic producers of X. This argument also applies if the subsidy were financed by nontrade taxes. The collection costs incurred may be quite high compared to the collection or administration costs of a protective tariff, and so might make the latter first-best. We have thus a qualification to the general argument that subsidies remain first-best in the presence of domestic divergences.

3,3 By-product Costs of Revenue-raising: The Need for 'Tradiiig-off The first-best way of dealing with domestic divergences as set out in the previous chapter will be affected by the by-product costs of raising revenue. This is true even though (subject to qualifications noted above) the principal message of the previous chapter is unaffected that domestic divergences should be corrected by subsidies (and taxes) as close as possible to the point of the relevant divergence. It will be necessary to trade off the gains from correcting the original divergence against the new by-product costs (distortion costs plus collection costs) that are imposed by the need to finance the subsidy. The subsidy should be just high enough for the marginal gain from partially correcting the divergence to be equal to the marginal by-product costs. Hence, as shown in Meade (I955a,

Subsidy Disbursement Costs

37

pp. 230-1), if the marginal by-product costs are positive, it will be optimal now not to make the subsidy so high that it corrects the divergence completely. The principle is the same as in the case of the second-best optimum tariff situation. But this time we are talking about a first-best situation since we are assuming that for every given degree of correction of the divergence by a subsidy, the byproduct costs from raising the revenue to finance the subsidy are minimized. One would normally expect the marginal by-product costs of raising revenue to be positive and so some trading-off to be necessary. This is true even though there are likely to be indivisibilities in tax collection costs, so that the marginal collection costs could indeed be zero over a range once the tax system has been set up, since costs may not increase significantly when tax rates are increased. But marginal distortion costs-- resulting from the disincentive effects and other distortions created by taxes—are likely to be positive, and indeed to increase with the amount of revenue raised.3 How is the hierarchy of policies described in the previous chapter affected by the by-product costs of raising revenue? At every stage in the hierarchy the gains to be derived from a particular policy will be less than in the absence of these costs. When a subsidy is used there will always be some collection and some distortion costs to be added to the by-product distortion costs already taken into account when setting up the hierarchy, When a tariff is used there will be tariff collection costs that were not taken into account in the previous chapter, as well as a disincentive effect (distortion relative to leisure) resulting from the consumption tax aspect of the tariff. Furthermore, it seems reasonable to argue that the order in the hierarchy will probably be unaffected. Policies at the top of the hierarchy are those which are directed precisely to the point of the divergence; relevant subsidies required will then cost rather little, less than when the subsidies are less discriminating. Hence the welfare gains from choosing a policy high up in the hierarchy as compared with one lower down will probably be even greater than before.

3,4 Subsidy Disbursement Costs The fiscal transaction of raising revenue and then using it for subsidization has two transaction costs, the collection cost of raising the revenue, and the cost of disbursing the subsidy. So far we have only allowed for the former. In a country with a developed taxation system embracing all sections of the community, subsidies can be provided through reductions or exemptions in taxes already payable. Net subsidies could be paid out through the same system. The more the subsidies overlap with an existing category of taxpayers, ^ One could envisage a situation where it would be optimal not to trade off but to make a full correction of the divergence even though the marginal by-product distortion cost is positive, This would require the marginal gain curve to have a kink and drop vertically to zero. We shad generally ignore the possibility of such a 'corner solution' here.

38

Four Assumptions of Theory

the lower the disbursement costs. There are always costs in complicating any system, or in setting up a new system, but in general, in developed countries subsidy disbursement is not likely to be very costly. On the other hand, subsidy disbursement can be costly in a developing country, and sometimes impossible. Just as there are untaxable sectors, so there are unsubsidizable sectors. How would one provide a production subsidy to handicraft manufacturers? An all-embracing subsidy on employment in manufacturing could also be administratively very difficult. If one wished to replace the existing protection of manufacturing industry provided by tariffs and import restrictions with a system of direct subsidization, one would, in many countries, have to disburse very high subsidies; it would not be sufficient to exempt industries from existing excise taxes, since these tend to be much lower and to be confined to a few products. In practice, the simplest form of subsidization is often through the provision of low-cost facilities by the state, or subsidization of some inputs that are produced by a few domestic firms or that are imported. This gives rise, of course, to by-product distortion in factor use. Thus subsidy disbursement costs increase the by-product costs of intervention, may make some forms of intervention impossibly costly, and may alter the order in the hierarchy of policies. They may make it costly to intervene precisely at the point of a marginal divergence because of the costs of defining and identifying the relevant recipients. Indeed, absolute precision in this regard is almost impossible, so that some by-product distortion costs resulting from an intervention would be inevitable even if there were no by-product costs of raising the revenue needed to finance the subsidy. One form of protection has no disbursement costs. The restriction of imports through a tariff or a quota causes the domestic price level of competing products to rise, so that automatically domestic consumers are taxed and producers subsidized. The simplicity of this device when compared with taxing, numerous domestic consumers and then disbursing the revenue to numerous producers is obvious. Exactly the same effect can be obtained by an export tax if it is desired to subsidize some consumers. The tax lowers the price level at which the product is exchanged domestically and hence it taxes producers and subsidizes domestic consumers in one action.4 An export subsidy may also be preferred to a production subsidy for this reason. It is often easier to disburse a subsidy at the port, when a product leaves the country or when the foreign exchange transaction takes place, than at the point of production. Furthermore, the amount to be disbursed is less than with a production subsidy if a given degree of protection is to be provided: the export subsidy causes the price of the exportable product sold to domestic consumers to rise (for otherwise producers would export all their production); hence there is no need to subsidize that part of production which is sold at home. 4 The tariff or the export tax will, or may, raise customs revenue. It might be argued that this will need to be disbursed and hence disbursement costs will be incurred. But this is not so. It can go into the general funds of the government, permitting other distorting taxes to be reduced.

Income Distribution Effects

39

We have thus an argument for protection through taxing or subsidizing trade rather than domestic production, consumption or factor use. It becomes conceivable that a tariff may be first-best on these grounds. This new argument must be set against all our earlier arguments on the other side. The force of the argument depends, of course, completely on the empirical question of whether subsidy disbursement costs are high. In developed countries they are unlikely to be high, nor in developing countries when it is a matter of protecting manufacturing industries in the advanced sector of the economy.

3.5 Income Distribution Effects and the Theory of Domestic Divergences Many economic theorists have tended to be either blind or nihilistic about the income distribution implications of economic policies, A vast amount of normative economic theory simply ignores income distribution by concentrating on 'Pareto-optimality.* The theory of domestic distortions on which Chapter 2 is based comes into this category. Some theorists, on the other hand, consider that every policy will have some effect on income distribution, that such effects cannot be ignored, that economists can make no interpersonal comparisons of utility, that therefore they cannot make any judgment about the desirability of any policy at all, and hence they cannot say anything useful at the normative level. On this basis the discussion in Chapter 2, and no doubt most a priori economic theory, would be irrelevant for practical issues. We shall be discussing these matters further in Chapter 5. Here we are concerned solely with how income distribution effects bear on the theory of domestic divergences. It will be argued that it is necessary neither to be blind nor to be nihilistic. In Chapter 2 we assumed that the raising of revenue involved no by-product distortion or collection costs and the disbursement of subsidies no disbursement costs. This meant that it would also be possible to alter income distribution by taxing some people and subsidizing others without any costs of this kind. Hence the assumption that income redistribution is costless would be consistent with this model. Let us make this assumption now. It will be removed in the next section. In any given situation, with any given set of marginal divergences, whether corrected or not, there will be (a) an actual income distribution resulting from that situation in the absence of any deliberate redistribution, and (b) a socially desired income distribution. We need not discuss the latter concept at this stage. Let us just accept that there is some kind of social ordering of possible income distributions. If the actual distribution differs from the socially desired one, then there is a domestic divergence. We can call it an income distribution divergence. This is very similar to the divergences we have discussed earlier.

40

Four Assumptions of Theory

First-best policy calls for an intervention as close as possible to the point of the divergence, namely appropriate subsidies and taxes on incomes. If some other device were used to redistribute incomes—for example a subsidy on production of some industries or a tariff- by-product distortions of the usual type would be generated, so that these policies would be second-best, third-best, and so on. Taxes and subsidies on trade, or for that matter on production or consumption, will never (given our assumptions) be first-best to deal with income distribution divergences. Now let us introduce the concept of the by-product income distribution distortion. Suppose that a marginal divergence caused by an externality is corrected by a first-best subsidy. This alters income distribution. Suppose, further, that the income distribution initially was the socially desired one and that the income distribution that results when the externality divergence has been corrected is not desired. The correction of the externality divergence has yielded a by-product income distribution distortion. It is then necessary to correct this with an appropriate income redistribution policy. Given our present assumption of costless income redistribution, it is possible to do this without creating any other by-product distortions. Hence the first-best package of policies involves the use of two instruments, (a) the first-best subsidy (say, a production subsidy) that corrects the initial divergence created by the externality, and (b) the income tax-subsidy that corrects the income distribution distortion created by the first subsidy. The two policies are thus directed toward two targets, (i) correction of the initial marginal divergence, and (ii) avoidance of an income distribution distortion. With two costless instruments both targets can be attained. This approach assumes, it must be repeated, that the raising of revenue involves no by-product distortion or collection costs and the disbursement of subsidies no disbursement costs. If we assume that income distribution distortions created by various policies are always corrected by appropriate first-best and hence costless income taxes and subsidies, we can ignore the income distribution consequences of different policies. The hierarchy of policies to deal with a wage-rate divergence, for example, will remain exactly as set out in Chapter 2.

3.6 Income Redistribution Not Costless The inevitable by-product distortion costs of raising revenue must now be reintroduced; hence income redistribution can no longer be costless. This requires the preceding discussion to be modified in three ways. In each case we suppose that there is initially the familiar marginal divergence in production owing to an external economy created by an import-competing industry, First, optimal subsidization policy must now be influenced by the income distribution effects. Trading-offis required. The subsidy to the industry concerned will increase the incomes of (a) the producers of the protected products

Income Redistribution

41

and (b) the persons who benefit from the external effect generated by the industry. If this redistribution is not in a desired direction, some redistributive policies are required. These will tax the beneficiaries in some way that does not reverse the initial effect of the subsidy on production, but will inevitably impose by-product distortion costs. It follows that optimization requires a trading-off process. The subsidy should not be so high as to correct the initial marginal divergence as far as it would have been corrected in the absence of an undesirable income distribution effect. Similarly, the income redistribution policy should not be fully corrective: some of the income distribution effect of the subsidy should stick. To assess the desirability of the optimum subsidy and how far it should go, it is thus necessary to take the income distribution consequences into account. One can no longer confine oneself to 'Pareto-optimality.* Second, a subsidy to correct the marginal divergence remains preferable to a tariff. So the main message of the theory of domestic divergences stands (in the absence of subsidy disbursement costs). Indeed, the argument in favor of a subsidy is strengthened. There are two reasons for this, (1) A subsidy redistributes incomes from the general taxpayers-••••••the community choosing to distribute the tax burden as it wishes—to the producers of the particular product. Income redistribution policy will try, at a cost, to offset, partially, the latter effect. By contrast, a tariff redistributes incomes from particular consumers to particular producers. Income redistribution policy has then a bigger job to do—to reverse, partially, not only the effect on the particular producers, but also the effect on the particular consumers. It may be that the income distribution was initially not optimal, and some redistribution against these consumers was desirable; it is still unlikely that the tariff would be the best way of achieving this.5 (2) The tariff raises customs revenue, which transfers income from particular consumers to the general taxpayers. But they may not be the particular consumers whom it is optimal to tax for income distribution reasons; it would be pure chance if this were so. Hence, in principle, some redistributive policies are needed to offset this. The tariff has thus a bigger income distribution effect than the subsidy, and if this effect is generally not in a desired direction, the required correction gives rise to more by-product costs. The fact that income redistribution is not costless has a third effect. Suppose the country is constrained to use a tariff, rather than a subsidy, for protection, so that we have the second-best situation described in Section 2.4. Suppose, further, that the income distribution consequences of the tariff do not happen, by chance, to be desirable. It must now be remembered that, even though income redistribution through income taxes and subsidies should accompany the tariff, it should not go so far as to restore the original distribution; because of the by-product costs of redistribution, some of the distributive effects of the 5

This argument will be qualified in Section 5,4,

42

four Assumptions of Theory

tariff should stick. It follows that the height of the second-best optimum tariff will be less than otherwise, and the less desirable the income distribution consequences of the tariff, the lower it should be.

3.7 Illusions and an Imperfect World How can subsidies always be preferable to trade interventions for correcting a domestic divergence, at least aside from possibly minor considerations of subsidy-disbursement costs and of particularly low collection costs of protective tariffs relative to revenue tariffs? Instinctively one rebels against so sweeping a generalization. Furthermore, the whole 'subsidy-biased' approach seems to have a great air of unreality about it. How does it corne about that tariffs and import restrictions are used all round the world when their use for what are often the professed purposes of the restrictions can be so broadly condemned? Surely one can hardly conceive of a country such as India, for example, replacing its regime of tariffs and import controls with a mass of subsidies to producers. What then have we left out? One cannot necessarily assume that the revenue to finance subsidies is always raised in a minimum-cost way. Of course it could be, in principle, but it just may not be. One has then the typical second-best problem of an imperfect world; one must compare a tariff, which is certainly second-best or worse, with a subsidy financed in a second-best way. In many countries it is always difficult politically to increase tax rates, though there is no similar resistance to tariffs, An imbalance between private spending and public facilities may result: we start then with a marginal- - indeed, much more than marginal—domestic divergence. A subsidy may well be financed not by increasing taxes but by forgoing other public expenditures and so increasing the existing imbalance. Similarly, one cannot necessarily assume that first-best income distribution policies are being followed, so that we do not start with an 'undistorted' income distribution situation. The protected products may, broadly, be 'luxury' goods consumed mainly by the middle and upper classes and the income distribution consequences of protection may therefore be desirable. If subsidies were used, the required revenue might be obtained by general sales taxes regressive in their effect, or perhaps by forgoing government expenditures on social services. This situation means that there is an inconsistency. If the people that control the government are willing to accept higher prices for imported luxury goods, why cannot the subsidy be financed by a sales tax on luxury goods, or an income tax on the sort of people who consume these goods? The answer is that people and governments are often inconsistent, or apparently so. Perhaps one interest group controls the tariff policy and another is able to influence tax

Illusions and Imperfect World

43

policies. More important, there may be an element of illusion. This is a matter of great importance for tariff and import quota policies, and needs to be looked at more closely. Explicit taxation imposes psychic costs which implicit taxation does not. These are costs in the minds of the taxed public and hence costs to the government in loss of popularity. One cannot ignore an element of irrationality and illusion in popular attitudes to taxation. The payment of personal income tax is a clear payment and is just as clearly resented, explicit taxes on production are noticed and resented by producers, while consumption (sales) taxes may be less obvious but are often brought to public notice by producers or retailers, though they are probably more acceptable to the public than income taxes. But tariff and import quotas, provided they are not mainly on inputs into domestic production, and especially if they are prohibitive, are another matter. It may be understood that on the imports that remain, tariffs are equivalent to sales taxes, but it is often not understood that the subsidy equivalent of the tariff is like a subsidy to producers financed by a sales tax. So a fiscally hardpressed government's reluctance to replace protective tariffs with subsidies is understandable. This is the cosmetic attraction of tariffs. A similar cosmetic attraction attaches to export taxes in countries where a high proportion of exportables is consumed domestically. It may be realized that the actual export tax revenue is collected, in effect, from producers. But it is not widely understood that the lower domestic price created by the export tax is a way of taxing producers of exportables for the benefit of consumers or of other industries. A subsidy financed by tax revenue and hence going through the budget makes it obvious that an industry is protected. A clear sum of money stands witness to the cost, even though it is not, strictly, a measure of welfare cost. By contrast, a tariff or quota hides the reality. Quite elaborate research may be needed to calculate the subsidy equivalent of the tariff or quota. For the same reason, if subsidies are chosen, the beneficiaries often prefer indirect subsidies given to factors of production or through the provision of public facilities below cost rather than direct output subsidies. Protection unnoticed is protection more secure. And, even when it is noticed, tariff legislation is not normally annually renewable, while by contrast, in democracies budgets are annual and their contents are often scrutinized with embarrassing thoroughness. For this very reason—the obscurantist aspect of tariffs and quotas—free-trade-minded economists preferred subsidies to tariffs long before the theory of domestic distortions was developed. The chances of sustained protection are certainly less with subsidies than with tariffs, so that it is in the interests of exporters and relevant consumers that explicit subsidies rather than tariffs are used as protective devices. At the same time, it must be noted that many countries do use subsidies to sustain very expensive activities that most economists would regard as uneconomic—as witness Britain and France's aircraft extravaganzas (the development of the Concorde, the

44

Four Assumptions of Theory

supersonic aeroplane)—but this is usually when tariff's could not achieve the protective purpose, as in the case of export industries. It is thus not difficult to understand the widespread use of tariffs and quotas in preference to subsidies, and perhaps in some cases, even to defend such use. But insofar as economists have a bias against protection on the grounds that generally it results from special interests overruling general interests, and insofar as they have a bias in favor of explicitness and rationality-—a belief not only that the "cost-benefit' technique is their special skill, but also that its practice is a good thing—then one would expect them to prefer subsidies to tariffs. This is quite apart from the likelihood that the use of subsidies will minimize the byproduct distortion costs of protection.

4

Trade Taxes as Sources of Government Revenue Taxes on. foreign trade are substantial sources of revenue for governments of almost all developing countries. In at least twenty developing countries trade taxes in the late 1950s accounted for one-quarter or more of total government revenue,1 In a number of countries—including Indonesia, Burma, Sri Lanka (then Ceylon), Malaysia, Thailand, Nigeria, Ghana, and Colombia—the proportion was over 40 percent. The main type of trade tax has been the tariff, but in addition there were the profits of multiple exchange rate systems, export taxes, and profits from export marketing boards, the latter being really forms of export taxes, Export taxes, including the profits of marketing authorities, were important sources of revenue in the early 1950s, though they declined in importance since then. Export marketing boards were established in a number of countries during or immediately after World War II to stabilize internal prices and improve marketing facilities, and rather incidentally yielded large revenue surpluses until the mid-1950s. Ordinary export taxes became important as a result of the Korean raw materials boom in 1950 51, when rates of export tax were increased in many countries. By the early 1960s there were twelve countries where export taxes (including marketing board profits) accounted for more than 10 percent of government revenue, and four (Uganda, Malaysia, Sudan, Sri Lanka) where the proportion was more than 20 percent,2 In developed countries, by contrast, taxes on trade were not significant sources of revenue by the 1970s, and usually accounted for less than 10 percent of central government revenue, and sometimes much less. But this was not always so. In the early histories of the now-developed countries trade taxes have often been very important, and the principal purpose of tariffs has been to raise revenue. In the United States, customs duties accounted for over 25 percent of revenues at all levels of government in 1890, though only 0,8 percent in 1960, In Germany they accoun ted for 16 percent of revenue in 1914, though only 4 percent in 1960 (Musgrave, 1969, pp. 138-9). Furthermore, in the 1 Lewis (1963), (Indonesia has been added to his list here.) This was the only available comprehensive statistical study of this subject by 1970. Subsequently Greenaway and Milner (1991) have provided figures for the average of the seven-year period 1976- 82, For nineteen developing countries (eleven in Africa), the share of trade tax revenue in total government revenue was 33 percent or greater, ' Goodeetal.(I966).

46

Trade Taxes as Revenue Sources

European Union countries revenue from customs duties which supplement taxes on domestic production, such as excise taxes or value-added taxes, are still of some significance. Such tariffs are 'border-tax adjustments* which convert taxes on production of import-competing goods into taxes on domestic sales. In view of the role that trade taxes play currently in developing countries as sources of government revenue, and that they once played in the now advanced countries, it is clearly necessary to consider trade taxes not only from a protective but also from a revenue point of view. The aim of this chapter is primarily to consider two issues. First, what role, if any, should taxes on trade play in an optimum revenue-tax package, bearing in mind the availability of other taxes, notably income, sales, and excise taxes? Second, given that trade taxes are used to raise revenue, what considerations determine the optimal structure of trade taxes for this purpose?

4,1 Some Simple Principles of Optimum Taxation The problem of the revenue role of trade taxes will be posed here in a limited way: what is the optimum way of raising a given amount of revenue, and especially what is the role of trade taxes in such an optimum system? We shall assume here that domestic divergences have all been corrected (though, in fact, it will normally not be optimal to make full corrections), that the country cannot influence its terms of trade, and that the exchange rate maintains external balance and monetary policy maintains internal balance. The revenue raised from taxes may be used to finance public goods, to finance subsidies to correct domestic divergences, to finance social services, or to finance capital accumulation. Illusion or 'cosmetic' aspects of various taxes will be ignored. We shall also assume throughout this chapter that savings are unaffected by the choice of tax, since any possible effects on savings are peripheral to the main issues here. Three considerations will then determine the optimal tax mix, namely income distribution effects, distortion effects, and collection costs. Collection costs will turn out to be crucial, for without them trade taxes cannot be part of a first-best tax package. Even if there were collection costs for income taxes™ perhaps even so high that income taxes were completely ruled out so that all revenue had to be raised with commodity taxes—it would remain true that trade taxes as such should not be used. One has to introduce collection costs not only for income taxes but also for commodity taxes to get a first-best case for revenue tariffs or export taxes. The argument can be developed in a number of stages. Suppose, to begin with, that there are no collection costs for any type of tax, and furthermore that the elasticity of substitution of effort for leisure for all potential taxpayers is zero, so that taxation has no disincentive effects. All revenue should then be raised by taxes on persons—on their wealth, incomes, or expenditures—and not on commodities, The issue of the choice between

Principles of Optimum Taxation

47

wealth, income, and expenditure taxes hinges on the effect on savings, which we are ignoring, and it will be simplest to assume an income tax. The desired income distribution effects can be brought about in a straightforward manner through the income tax. One can apply any desired principle for distributing the burden of taxation—ability to pay, benefit principle, class interest, and so on. To avoid any distortion the tax must not be related in any way to the type of activity or industry in which taxpayers is engaged or to the way in which they spend their incomes. When we allow for the effects of taxation on incentives, complications are introduced which have long preoccupied economists. There are then inevitable distortion costs because taxation brings about substitution of the untaxable good, leisure, for effort, the reward of the latter being taxed income. Given that there are inevitable distortion costs, these can be reduced by modifying the simple income tax system based on income distribution considerations alone. First, progressiveness can be reduced. Second, taxes can be raised relatively on those members of the community who have relatively less opportunity to vary the supplies of their factor services. Finally, an element of differential commodity taxation can be superimposed on the income tax system, so that prices of goods that are poor substitutes for leisure, or are actually complementary to leisure, rise relatively to the prices of goods that are close substitutes for leisure.3 These adjustments will reduce somewhat the distorting substitution toward leisure that taxation is likely to bring about, though the resultant gains must be traded off against the failure to attain the desired income distribution objectives and against the new distortion in the consumption pattern between the various goods concerned that is introduced. The last adjustment, concerned with various goods being substitutes for, or complementary to, leisure to different degrees, is probably unimportant, and it will henceforth be assumed that all goods are equal substitutes for leisure. The next step is to suppose that collection costs for income tax are so high that all revenue must be raised from taxes on commodities—sales taxes, excise taxes (taxes on domestic production), and trade taxes. We might, alternatively, suppose that some income tax is being collected, but increases in income tax are not possible. In the absence of disincentive effects, distortion costs would now be avoided by a uniform ad valorem tax either on domestic production or on domestic sales. If there are disincentive effects but if all goods are equal substitutes for leisure, a uniform tax would minimize, rather than avoid, distortion costs. But a uniform tax does not make it possible to take into account income distribution considerations since it is equivalent to a proportional income tax. If it is desired to impose the tax burden mainly on the rich, it becomes appropriate to have a system consisting of a differential sales tax combined with a differential excise tax. The sales tax structure would consist of 3

Corlett and Hague (1953-54), Meadc (!955a, pp. 112 18), and Meade (I955b. pp. 29- 30).

48

Trade Taxes as Revenue Sources

higher taxes on luxury' goods consumed by the rich and lower taxes on 'essentials' consumed by the poor. The excise tax structure would consist of higher taxes on those products which are intensive in factors earning high incomes, and low or zero taxes on products intensive in low-income factors. While there is still no role for trade taxes as such, a trade tax could be a component of either a sales tax or an excise tax. A sales tax on an import-competing good could be levied in the form of a tariff combined with a tax on domestic production at the same rate (with an allowance for wholesale and retail trading margins). Similarly, a production tax on an exportable good could be levied in the form of an export tax combined with a tax at the same rate on sales of the exportable product to domestic consumers. Finally, we can narrow down the range of taxes even more by supposing that high collection costs compel the exclusion not only of income tax but of all kinds of taxes except two, namely (a) tariffs and (b) excise taxes on domestic production of import-competing goods. We can also suppose that a sales tax on importables is possible since this is more or less equivalent to a combination of tariff and excise tax. The assumptions are somewhat extreme, but in many developing countries income tax can be levied only in the advanced sector of the economy, and this sector produces mainly import-competing goods, so that the income tax is essentially like a production or excise tax on importcompeting production. Since export taxes are (broadly) symmetrical with tariffs, one could also introduce export taxes without altering the analysis significantly. The problem then arises of making up an optimum mix of taxes. Essentially the choice is between a tax on import-competing production and a tax on import-competing consumption, the latter tax being obtained by combining a tariff with an excise tax at the same rate. Each form of tax creates a distortion, and the two distortions must be traded off against each other. In addition, income distribution must be taken into account: a shift from production tax to consumption tax would shift the burden of taxation away from importcompeting producers toward exporters, and toward those who are especially heavy consumers of importables. There is still no first-best role for a tariff, unaccompanied by an excise tax on domestic production. This point can be made clearly if we imagine, realistically, that it is desired to raise a given amount of revenue from commodity taxes on importables, the burden of the tax to be borne mainly by the richer members of the community. It can be shown that a tariff on luxury' goods on its own— unaccompanied by an excise tax would be second-best in this case; it would involve a greater distortion cost for given revenue raised than a sales tax, and in addition would involve by-product income distribution effects in, favor of domestic import-competing producers which would be in a desired direction only by pure coincidence. This comparison between a sales tax and a tariff is made in Figure 4.1. It refers to an importable good consumed mainly by rich people. DD' is the

Collection Costs

49

FIG. 4.1

domestic demand curve, Hff the import-competing supply curve, and SS' the foreign import supply curve, A sales tax of 57* would raise revenue of STLJ and inflict a consumptiondistortion cost of JLG. It would be the equivalent of a tariff of SFcombined with an excise tax on domestic production, also of ST. A tariff on its own to raise the same revenue would have to be 57*, raising revenue of KK'N'N (=STLJ). It would inflict the production-distortion cost RK A" and the consumption-distortion cost NN'G. Thus, not only would it add a production-distortion cost, but the consumption-distortion cost would be greater than in the case of the sales tax because the rate of tax would have to be greater to raise the same revenue. In addition, it would bring about a redistribution of income not only from consumers of the product to the Treasury equal to the revenue, but also (in partial equilibrium terms) a redistribution of STK'R from these consumers to import-competing producers; and the latter redistribution might not be socially desired, for example, if the producers were even richer than the particular consumers, or if the producers were foreigners, It follows that only relatively high collection costs of sales or excise taxes can justify a preference for tariff over sales tax.

4.2 The Importance of Collection Costs Economists have tended to ignore collection costs of taxes, and in general have neglected the administrative aspects of fiscal systems.4 Any particular economist can be excused from analyzing these matters in detail when they 4 There are exceptions. See especially Due (1970), which is the best reference on the subject. Also Shoup et al, {1959) and Musgrave (1969), Johnson {1965) dismisses administrative costs as 'of practical rather than theoretical consequence,' and similarly, these matters ind no place in either of the two modern classical theoretical studies, Meade (I955a) and Musgrave (1959).

50

Trade Taxes as Revenue Sources

happen to be outside his or her personal area of competence; for this reason they are not discussed in detail in this book. But one cannot excuse a failure to stress their importance, and to build them into theoretical models, when these models are meant to be some guide to the choice between different types of tax, or between complicated and simple versions of any particular type of tax. The costs with which we are concerned are really of three types. First, there are the costs of tax administration incurred by the tax-collecting authority; an important element of this is the labor cost of ensuring compliance. Second, there are the resource costs incurred by taxpayers to fulfill their tax obligations while minimizing payment, or perhaps to ensure successful evasion; these costs can be quite considerable when complicated taxes are involved or when taxpayers actually choose to break the law. Bribery of tax officials should not be included in these costs, since bribes as such are simply ineomeredistributive. Third, there are the distortion costs which result from taxpayers rearranging their affairs as part of an avoidance or evasion effort. In principle these are no different from the rearrangements of production and consumption patterns, or of the choice between work and leisure, which have already been allowed for in our analysis. But tax systems are often designed to minimize such distortions, and the less effectively a tax system is administered, the more distortions one can expect. The term collection costs will now be used to embrace all three types of cost, even though a part of the second cost and the whole of the third cost are attributable to taxpayers' attempts to reduce collections, and we will mainly have in mind the first type of cost, namely the cost of tax administration. The central point is that collection costs for trade taxes are generally much lower than for other taxes. This is mainly because foreign trade usually flows through a few ports or bottlenecks, and even when it does not, it is easier to police a border and collect taxes on goods passing across it than to seek out a large number of individual taxpayers, whether persons or firms, or to ensure that they produce accurate tax returns. The point is really quite obvious. In developing countries collection costs of export taxes are sometimes especially low, lower indeed than costs of import taxes, both because exports are often more homogeneous products than most imports, and because, in many cases, there are only a few main products and—even more conveniently—a few main producers. Once we allow for collection costs, tariffs and export taxes may form part of a first-best tax package. The tariffs need not necessarily be accompanied by excise taxes on domestic production at the same rate. This is indeed the reason why taxes on trade are so significant in developing countries. In determining the optimum tax mix, say, between tariffs and excise taxes, differential collection costs may weight the scales heavily in favor of tariffs in spite of the production distortion cost that may result from the protective effects of tariffs.

Smuggling and Under-Invoicing

51

Collection costs also present an argument for indirect taxes collected on domestic production (like an excise or a value-added tax) or on wholesale or retail sales, in preference to a comprehensive personal income tax, since the indirect taxes require contact with far fewer taxpayers. In some countries a case for a corporate income tax might also be based on collection costs. In most developing countries income tax is collected only from very few persons and from large companies, partly because of collection costs, and hence reliance has to be placed on indirect taxes. In an economy with a high ratio of imports to total domestic sales, the ease of collecting tariffs on imports is a strong argument for the use of sales taxes, or taxes having effects similar to sales taxes, since a substantial part of these can be collected at the point of importation. In comparing tariffs with more general indirect taxes on domestic sales, one must compare the costs of collecting taxes on imports with the costs of taxing domestic production. Most developing countries do have taxes on domestic production, whether described as excises or sales taxes, but they usually apply to only a limited number of products produced or sold by large establishments. The problem is the difficulty of collecting taxes from small-scale establishments. For this reason, a retail sales tax is usually not practicable. It may be technically possible to collect taxes from small firms, so that they are not literally 'untaxable,' but the costs in relation to tax yield would be very high.

4.3 Smuggling and Underinvoicing Collection and evasion problems also arise in the case of taxes on trade, though generally they are less critical than for other taxes. Essentially there are three problems: smuggling, underinvoicing, and misclassification.5 (1) Smuggling is the entry of unrecorded and hence untaxable imports, or the exit of unrecorded exports. It is a particular problem for island nations. Its prevention can use up substantial resources of a customs administration. Smuggling has been a serious problem for Indonesia and the Philippines, two multi-island states. Smuggling into and out of these countries is eased by the proximity of Singapore and Hong Kong, both great commercial centers with which traders in Indonesia and the Philippines have close business connections. Comparison of prices in Djakarta markets with Singapore and Hong Kong prices of comparable goods indicates that in many cases importers cannot have paid the legal tariffs. Smuggling is a form of tax evasion which involves resource costs for the smugglers, as well as risks of discovery and prosecution to which some monetary value could be attached, and smugglers are likely to balance these against the tariff or export tax payments avoided. Hence the lower the ad valorem rates 5 See Due (1970. pp. 38-53), Bhagwati (1969). Sitnkin (1970), and Richtcr (1970). I have also benefited from seeing an unpublished paper on the Indonesian tariff system by Richard Cooper.

52

Trade Taxes as Revenue Sources

of tariff or export tax and the greater the difficulty of smuggling—which depends, in particular, on the size and weight of the smuggled goods—the less smuggling there is likely to be, Wristwatches, on which a high ad valorem duty used to be payable, have long been the favorite smuggled item in Britain. When all imports are legal and hence taxable, there is one tariff rate which maximizes the customs revenue for any product, the level of this maximum revenue tariff being higher the lower the elasticity of the import demand curve for that product.6 Once we allow for smuggled imports, we can conceive of another demand curve, this time only for legal imports, which will be more elastic than the demand curve for imports of the product as a whole because its elasticity will take into account the falling ratio of legal to smuggled imports as the tariff rate risea The maximum revenue tariff rate will then be lower than if there were no smuggling. A similar argument applies on the export side: smuggling will lower the maximum revenue export tax rate. (2) Underinvoicing means that the values of imported or exported goods are invoiced by traders below the actual values paid or received by them. Hence, when tariff rates are fixed on an ad valorem basis (related to value of imports), there is a loss of revenue. The collaboration of foreign exporters may be required for this, and if they obtain some reward for their collaboration, the prices paid by importers will be higher than otherwise; so underinvoicing might worsen the terms of trade. Similarly, underinvoicing of exports leads to a loss of export tax revenue and some of the gains may go to foreign importers. The problem of obtaining correct values of imported goods for customs purposes is a very important aspect of customs administration. It is avoided when customs duties are specific rather than ad valorem, but specific duties have many disadvantages and inequities, and are not suitable for highly differentiated manufactured goods. Normally customs authorities deal with the problem by having lists of foreign trade prices from principal supplying countries, using, for example, manufacturers* catalogues. Evidence that there is ma/-invoicing (under- or overinvoicing) of imports can sometimes be obtained by comparing the import statistics of the country concerned with figures of exports from its supplying countries (since there may be no incentive to misrepresent export values in the supplying country). Similarly, evidence of underinvoicing of exports can sometimes be obtained by comparing recorded export figures with import statistics of customer countries. Underinvoicing of imports, like smuggling, means that the actual payment for imports made by importers is greater than that recorded in the import statistics; assuming that there is no underinvoicing of exports at the same time, underinvoicing of imports must show up somewhere else in the balance of payments as some form of unrecorded payment, remittance or capital outflow, or perhaps just as a residual debit item. But if there is underinvoicing also on the export side this may not be so; the foreign exchange unrecorded and retained 6

See section 4.4 below,

Marginal Cost of Revenue

53

by Indonesian exporters in Singapore may be sold directly to Indonesian importers, not passing through the Indonesian Central Bank, Hence both export and import tax payments are reduced and export and import values may be understated to an equal extent, In many countries restrictions on certa in imports and on capital outflow create an opposing incentive for twrinvotcing of imports. The purpose of overinvoicing is to obtain foreign exchange from the central bank nominally to buy legal imports, but in fact to sell part of this foreign exchange on the black market, where a high rate is sustained by restrictions on certain imports which can only be obtained with black market foreign exchange. Alternatively, the excess foreign exchange may be stowed away in Hong Kong or Swiss bank accounts. For the same reason there may be uruferinvoicing of exports, (3) Finally, another major problem of customs administration is that of misclassification of goods. This problem only arises because tariffs are not generally uniform ad valorem. If there is any doubt about the appropriate classification for a good, importers will have an incentive to make it appear to fit into a low-duty category. In a world of continually changing and differentiated products, classification systems need to be continually revised. The greater the costs of administration and the less developed the customs administration, the stronger the case for uniform ad valorem nominal tariffs, or at least for having only a small number of categories. Distinctions made on the basis of end use present particular problems: if an automobile for business use is taxed at a lower rate than one for private use, how can the customs inspector be sure that the 'business' automobile will not be resold for private use, or that business and private use will not be combined? Having said all this, it must be pointed out that customs administration is highly developed in most advanced countries. In most developing countries collection costs are significant and evasion of taxes on trade do present some problems (even in a country with such a highly developed administrative system as India), but the problem is only severe in a limited number of countries—notably the island states of Indonesia and the Philippines, and some African countries. Collection costs and evasion problems of trade taxes seem to be generally much less than for most nontrade taxes.

4.4 The Optimum Tariff Structure: The Marginal Cost of Raising Revenue If a given total revenue is to be raised from tariffs alone, what tariff structure will minimize the distortion costs of raising this revenue? This question is quite distinct from the question asked so far, namely whether tariffs, rather than other taxes, should be used to raise revenue. We now take the decision to raise revenue through trade taxes as given. It will be assumed that there are many importable goods, and hence many possible tariff rates, and no domestic

54

Trade Taxes ax Revenue Sources

divergences. The small country assumption applies, as usual. Income distribution effects are disregarded. Collection costs per dollar of revenue raised are assumed to be the same for all tariffs, and hence can be disregarded.7 If (a) the elasticity of supply of exportables and of domestic demand for exportables were zero and if (b) taxation had no disincentive effects, with the elasticity of supply of effort zero, the answer would be simple. Tariffs would not distort the production or consumption pattern relative to exportables or leisure, and the only distortion possible would be in the pattern of production and consumption of importables. The optimum tariff structure would be a uniform tariff, A uniform nominal tariff would avoid distortion of the consumption pattern, and a uniform effective tariff would avoid distortion of the production pattern. If exportables were never inputs in importable production, a uniform nominal tariff would automatically lead to a uniform effective tariff; otherwise, precise uniformity in both could not be attained, and some distortion would be inevitable (Corden, 197la, pp. 188-9). If substitution effects relative to leisure and to exportables have to be allowed for, the answer is different. The optimum revenue structure should then be nommiform, based on the principle familiar from public finance theory that, to minimize deadweight costs, taxes on low-elasticity goods should be higher than those on high-elasticity goods. The conclusion applies in a general equilibrium model, but a simple partial equilibrium exposition will be used here. We shall ignore the leisure substitution and focus on substitution relative to exportables. Suppose that for every importable product we can draw a demand and a supply curve which shows how quantities consumed and produced change as the price of the relevant product changes relative to the fixed prices of exportables; each of these curves is assumed to be independent of all the other curves, cross-elasticities being zero. For each product the supply curve is then subtracted from the demand curve, and a demand curve for imports is obtained. The optimum revenue- tariff structure will now involve high tariffs on goods where the elasticity of import demand is low—so that little distortion is caused by a tariff—and low tariffs on goods where the elasticity of import demand is high. It has to be remembered that the elasticity of import demand is a derived elasticity; it could be high because the good is a close substitute domestically for exportables either on the demand or the supply side. A useful concept is the marginal cost of raising revenue. This must be distinguished from ordinary marginal revenue. Figure 4.2 shows the import demand curve, DD', for one product, while SS' is the import supply curve. A tariff of ST raises revenue of STHG at a total distortion cost of GHJ", the latter is the sum of i\ie, production-distortion cast and the consumption-distortion cost. As the tariff is increased, the revenue increases up to the maximum revenue tariff 57*. The revenue STH'G' is the maximum 7

See Bliss (1992), and literature cited there, for further formal development of this topic.

Marginal Cost of Revenue

55

FIG. 4.2

attainable. Note that when revenue is at its maximum, ordinary marginal revenue is equal to the import price (MM' is the ordinary marginal revenue curve, and crosses SS' at (?')• As the tariff rate is increased up to the maximum revenue tariff, both revenue and the distortion cost increase. Hence there is a relationship between the distortion cost of raising the revenue and the actual revenue raised. This then yields a marginal cost of raising revenue. At the maximum revenue tariff rate, this marginal cost will be infinite.^ If the £>£)' curve becomes steeper (so that the elasticity of the curve falls at /), a given tariff rate will both raise more revenue and impose a lower distortion cost. This is illustrated in Figure 4.3 for the case where there are two imports and the two import demand curves are linear. The marginal cost of raising revenue is shown on the vertical axis, and the tariff rate on the horizontal. The curve relating the two for product I is